Summary
Turns thumbs-down feedback into executable prevention rules that block AI agents from repeating mistakes before the tool call executes. The PreToolUse hook intercepts risky actions like force-pushes, destructive file operations, or API calls that match patterns learned from prior sessions. One thumbs-down creates a rule, next session the agent gets blocked with zero tokens spent. Ships with a context brain generator that consolidates lessons, guardrails, and gates into a single agent-readable markdown file your repo can version in git. Free tier gives you unlimited feedback capture and five active rules. Works with Claude Code, Cursor, Gemini CLI, Cline, and any MCP-compatible agent. Designed for workflows where one repeated mistake is a liability event or a token bill you don't want to pay twice.
Plug Mailtrap into your AI workflow and let it handle the email.
Connect Mailtrap MCP →
Agent, run crypto. Access onchain data & trade routes via 1inch.
Install now →
Make coding agent sessions - Searchable, Shareable, Vendor-neutral & Scored.
Try For Free →
Your agent targets a perfect 10 Code Health score. Deterministic. Every commit.
Try For Free →
Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →
belt cli automatically finds the best tools and skills for your agent. image, video, music, tts...
one prompt install →Tools
Public tool metadata for what this MCP can expose to an agent.
68 tools
capture_feedbackCapture an up/down signal plus one line of why. Vague feedback is logged, then returned with a clarification prompt instead of memory promotion.13 params
Capture an up/down signal plus one line of why. Vague feedback is logged, then returned with a clarification prompt instead of memory promotion.
Parameters* required
chatHistoryarrayOptional caller-supplied recent conversation window used for history-aware lesson distillation. The current Claude auto-capture path sends up to 8 prior recorded entries for vague negative inline s...
contextstringOne-sentence reason describing what worked or failed
conversationWindowarrayRecent conversation turns before the feedback signal. Raw messages, not summaries.
failureTypestringDual-signal: "decision" = wrong tool/action chosen, "execution" = right tool but bad parameters/output. Improves Thompson Sampling precision.one of
decision · executionguardrailsobjectrelatedFeedbackIdstringOptional prior feedback event to merge with later follow-up context.
rubricScoresarraysignalstringone of
up · downskillstringtagsarraywhatToChangestringwhatWentWrongstringwhatWorkedstringfeedback_summaryGet summary of recent feedback1 params
Get summary of recent feedback
Parameters* required
recentnumbersearch_lessonsSearch promoted lessons and show the corrective actions, lifecycle state, prevention rules, gates, and next harness fixes linked to each result.4 params
Search promoted lessons and show the corrective actions, lifecycle state, prevention rules, gates, and next harness fixes linked to each result.
Parameters* required
categorystringone of
error · learning · preferencelimitnumberMaximum results to return (default 10)
querystringSearch query. Leave empty to list the most recent lessons.
tagsarrayRequire all tags to be present on a lesson
retrieve_lessonsRetrieve the most relevant lessons for a given tool/action context. Use in PreToolUse hooks for per-action guidance.3 params
Retrieve the most relevant lessons for a given tool/action context. Use in PreToolUse hooks for per-action guidance.
Parameters* required
actionContextstringDescription of what the tool call is doing
maxResultsnumberMax lessons to return (default 5)
toolNamestringThe tool being called (e.g., Bash, Edit, Read)
search_thumbgateSearch raw ThumbGate state across feedback logs, ContextFS memory, prevention rules, and imported policy documents.4 params
Search raw ThumbGate state across feedback logs, ContextFS memory, prevention rules, and imported policy documents.
Parameters* required
limitnumberMaximum results to return (default 10)
querystringSearch query for ThumbGate state.
signalstringOptional feedback-signal filter when searching feedback data.one of
up · down · positive · negativesourcestringRestrict search to a single ThumbGate source.one of
all · feedback · context · rules · documentsimport_documentImport a local policy or runbook document into ThumbGate, normalize it for search, and propose provenance-backed gate candidates.7 params
Import a local policy or runbook document into ThumbGate, normalize it for search, and propose provenance-backed gate candidates.
Parameters* required
contentstringInline document content for hosted or generated imports.
filePathstringLocal file path inside the active workspace or ThumbGate runtime.
proposeGatesbooleanWhen true (default), derive reviewable gate proposals from the document.
sourceFormatstringOptional source format override when importing inline content.one of
markdown · text · yaml · json · htmlsourceUrlstringOptional external URL or provenance label for the imported document.
tagsarrayOptional tags such as policy, runbook, or team.
titlestringOptional display title override.
list_imported_documentsList imported policy and runbook documents stored in local ThumbGate state.3 params
List imported policy and runbook documents stored in local ThumbGate state.
Parameters* required
limitnumberMaximum documents to return (default 20).
querystringOptional title or excerpt filter.
tagstringOptional tag or matched template id filter.
get_imported_documentRead a previously imported document with its proposed gate candidates and provenance.1 params
Read a previously imported document with its proposed gate candidates and provenance.
Parameters* required
documentIdstringImported document id.
feedback_statsGet feedback stats and recommendations
Get feedback stats and recommendations
No parameter schema in public metadata yet.
diagnose_failureDiagnose a failed or suspect workflow step using MCP schema, workflow, gate, and approval constraints.13 params
Diagnose a failed or suspect workflow step using MCP schema, workflow, gate, and approval constraints.
Parameters* required
approvedbooleancontextstringerrorstringexitCodenumberguardrailsobjectintentIdstringmcpProfilestringoutputstringrubricScoresarraystepstringtoolArgsobjecttoolNamestringverificationobjectinfer_lesson_from_historyPerform autonomous inference on chat history to identify why a failure occurred and what rule should be recorded.2 params
Perform autonomous inference on chat history to identify why a failure occurred and what rule should be recorded.
Parameters* required
chatHistoryarraylastActionobjectlist_intentsList available intent plans and whether each requires human approval in the active profile3 params
List available intent plans and whether each requires human approval in the active profile
Parameters* required
bundleIdstringmcpProfilestringpartnerProfilestringplan_intentGenerate an intent execution plan with policy checkpoints8 params
Generate an intent execution plan with policy checkpoints
Parameters* required
approvedbooleanbundleIdstringcontextstringdelegationModestringone of
off · auto · sequentialintentIdstringmcpProfilestringpartnerProfilestringrepoPathstringstart_handoffStart a sequential delegation handoff from a delegation-eligible intent plan9 params
Start a sequential delegation handoff from a delegation-eligible intent plan
Parameters* required
approvedbooleanbundleIdstringcontextstringdelegateProfilestringintentIdstringmcpProfilestringpartnerProfilestringplannedChecksarrayrepoPathstringcomplete_handoffComplete a sequential delegation handoff and record verification outcomes8 params
Complete a sequential delegation handoff and record verification outcomes
Parameters* required
attemptsnumberhandoffIdstringlatencyMsnumberoutcomestringone of
accepted · rejected · abortedresultContextstringsummarystringtokenEstimatenumberviolationCountnumberdescribe_reliability_entityGet the definition and state of a business entity (Customer, Revenue, Funnel). Aliased to describe_semantic_entity.1 params
Get the definition and state of a business entity (Customer, Revenue, Funnel). Aliased to describe_semantic_entity.
Parameters* required
typestringone of
Customer · Revenue · Funnelget_reliability_rulesRetrieve active prevention rules and success patterns. Aliased to prevention_rules.
Retrieve active prevention rules and success patterns. Aliased to prevention_rules.
No parameter schema in public metadata yet.
enforcement_matrixShow the full Enforcement Matrix: feedback pipeline stats, active pre-action gates, and rejection ledger with revival conditions.
Show the full Enforcement Matrix: feedback pipeline stats, active pre-action gates, and rejection ledger with revival conditions.
No parameter schema in public metadata yet.
security_scanScan code for OWASP vulnerabilities (injection, XSS, path traversal, SSRF, prototype pollution) and supply chain risks (typosquatting, install script abuse, wildcard versions). Returns findings with severity, category, and line numbers.3 params
Scan code for OWASP vulnerabilities (injection, XSS, path traversal, SSRF, prototype pollution) and supply chain risks (typosquatting, install script abuse, wildcard versions). Returns findings with severity, category, and line numbers.
Parameters* required
contentstringCode content to scan
diffModebooleanWhen true, treats content as git diff output
filePathstringFile path for language-aware scanning
capture_memory_feedbackCapture success/failure feedback to harden future workflows. Aliased to capture_feedback.3 params
Capture success/failure feedback to harden future workflows. Aliased to capture_feedback.
Parameters* required
contextstringsignalstringone of
up · downtagsarraybootstrap_internal_agentNormalize a GitHub/Slack/Linear trigger into startup context, construct a recall pack, prepare a git worktree sandbox, and emit an execution plus reviewer-lane plan.15 params
Normalize a GitHub/Slack/Linear trigger into startup context, construct a recall pack, prepare a git worktree sandbox, and emit an execution plus reviewer-lane plan.
Parameters* required
approvedbooleancommentsarraycontextstringdelegationModestringone of
off · auto · sequentialintentIdstringmcpProfilestringmessagesarraypartnerProfilestringprepareSandboxbooleanrepoPathstringsandboxRootstringsourcestringone of
github · slack · linear · api · clitaskobjectthreadobjecttriggerobjectprevention_rulesGenerate prevention rules from repeated mistake patterns2 params
Generate prevention rules from repeated mistake patterns
Parameters* required
minOccurrencesnumberoutputPathstringexport_dpo_pairsExport DPO preference pairs from local memory log1 params
Export DPO preference pairs from local memory log
Parameters* required
memoryLogPathstringexport_hf_datasetExport ThumbGate agent traces and DPO preference pairs as a HuggingFace-compatible dataset. Produces traces.jsonl, preferences.jsonl, and dataset_info.json with PII-redacted paths. Ready for huggingface-cli upload.2 params
Export ThumbGate agent traces and DPO preference pairs as a HuggingFace-compatible dataset. Produces traces.jsonl, preferences.jsonl, and dataset_info.json with PII-redacted paths. Ready for huggingface-cli upload.
Parameters* required
includeProvenancebooleanInclude provenance events in traces (default: true)
outputDirstringOutput directory (default: feedback-dir/hf-dataset)
export_databricks_bundleExport ThumbGate logs and proof artifacts as a Databricks-ready analytics bundle1 params
Export ThumbGate logs and proof artifacts as a Databricks-ready analytics bundle
Parameters* required
outputPathstringconstruct_context_packConstruct a bounded context pack from contextfs4 params
Construct a bounded context pack from contextfs
Parameters* required
maxCharsnumbermaxItemsnumbernamespacesarrayquerystringevaluate_context_packRecord evaluation outcome for a context pack6 params
Record evaluation outcome for a context pack
Parameters* required
guardrailsobjectnotesstringoutcomestringpackIdstringrubricScoresarraysignalstringcontext_provenanceGet recent context/provenance events1 params
Get recent context/provenance events
Parameters* required
limitnumbergenerate_skillAuto-generate Claude skills from repeated feedback patterns. Clusters failure patterns by tags and produces SKILL.md files with DO/INSTEAD rules.2 params
Auto-generate Claude skills from repeated feedback patterns. Clusters failure patterns by tags and produces SKILL.md files with DO/INSTEAD rules.
Parameters* required
minOccurrencesnumberMinimum pattern occurrences to trigger skill generation (default 3)
tagsarrayFilter to specific tags
recallRecall relevant past feedback, memories, and prevention rules for the current task. Call this at the start of any task to inject past learnings into the conversation.3 params
Recall relevant past feedback, memories, and prevention rules for the current task. Call this at the start of any task to inject past learnings into the conversation.
Parameters* required
limitnumberMax memories to return (default 5)
querystringDescribe the current task or context to find relevant past feedback
repoPathstringOptional repository path for structural impact analysis on coding tasks
unified_contextAssemble a complete, role-aware context object in one call. Combines session state, user profile, relevant lessons, prevention guards, context pack, and code-graph impact — with tiered graceful degradation (full → warm → cold). Replaces multiple recall/retrieve/session_primer...5 params
Assemble a complete, role-aware context object in one call. Combines session state, user profile, relevant lessons, prevention guards, context pack, and code-graph impact — with tiered graceful degradation (full → warm → cold). Replaces multiple recall/retrieve/session_primer...
Parameters* required
agentTypestringAgent type — shapes context budget and feature inclusionone of
claude · cursor · forgecode · codexquerystringDescribe the current task to find relevant context
repoPathstringRepository path for code-graph impact analysis
toolInputobjectCurrent tool input (for guard evaluation)
toolNamestringCurrent tool being invoked (improves lesson matching)
satisfy_gateSatisfy a gate condition with optional structured reasoning. Evidence is stored with a 5-minute TTL. When structuredReasoning is provided, the premise/evidence/conclusion chain is stored in the audit trail.3 params
Satisfy a gate condition with optional structured reasoning. Evidence is stored with a 5-minute TTL. When structuredReasoning is provided, the premise/evidence/conclusion chain is stored in the audit trail.
Parameters* required
evidencestringEvidence text (e.g., "0 unresolved threads")
gatestringGate condition ID to satisfy (e.g., pr_threads_checked)
structuredReasoningobjectStructured pre-gate reasoning: state premises, trace evidence, assess risk, derive conclusion before unlocking.
set_task_scopeDeclare or clear the current task scope so ThumbGate can compare affected files and diffs against the approved path set.7 params
Declare or clear the current task scope so ThumbGate can compare affected files and diffs against the approved path set.
Parameters* required
allowedPathsarrayGlob patterns that define the allowed file scope for this task
clearbooleanClear the current task scope instead of setting one
localOnlybooleanWhen true, also marks the task as local-only
protectedPathsarrayOptional protected-file globs that require explicit approval before editing or publishing
repoPathstringOptional repo root used when evaluating git diff scope
summarystringShort summary of the task being worked
taskIdstringOptional stable task identifier (ticket, issue, or work item id)
get_scope_stateReturn the active task scope and any unexpired protected-file approvals.
Return the active task scope and any unexpired protected-file approvals.
No parameter schema in public metadata yet.
set_branch_governanceDeclare or clear branch and release governance so PR, merge, release, and publish actions can be evaluated against explicit workflow state.11 params
Declare or clear branch and release governance so PR, merge, release, and publish actions can be evaluated against explicit workflow state.
Parameters* required
baseBranchstringProtected base branch for merge and release operations (defaults to main)
branchNamestringOptional branch name the governance applies to
clearbooleanClear the current branch governance state instead of setting it
localOnlybooleanWhen true, PR, merge, release, and publish actions are blocked for this lane
prNumberstringOptional pull request number once a PR exists
prRequiredbooleanWhether this lane must go through a pull request (defaults to true)
prUrlstringOptional pull request URL once a PR exists
queueRequiredbooleanWhether the target branch requires a merge queue
releaseEvidencestringOptional evidence or release plan note for the governed version
releaseSensitiveGlobsarrayOptional custom globs that define release-sensitive files for this branch lane
releaseVersionstringExpected package version for release or publish actions
get_branch_governanceReturn the active branch and release governance state.
Return the active branch and release governance state.
No parameter schema in public metadata yet.
approve_protected_actionGrant a time-limited approval for edits or publish actions that touch protected files.5 params
Grant a time-limited approval for edits or publish actions that touch protected files.
Parameters* required
evidencestringOptional supporting evidence or approval note
pathGlobsarrayProtected-file globs covered by this approval
reasonstringWhy this protected-file action is approved
taskIdstringOptional task id this approval is tied to
ttlMsnumberOptional approval lifetime in milliseconds (defaults to 1 hour, max 24 hours)
track_actionRecord a verification action in the current session (for example figma_verified or tests_passed). Session actions expire after one hour.2 params
Record a verification action in the current session (for example figma_verified or tests_passed). Session actions expire after one hour.
Parameters* required
actionIdstringVerification action ID to record
metadataobjectOptional structured metadata describing the evidence source
verify_claimCheck whether a claim has enough tracked evidence before the agent asserts it.1 params
Check whether a claim has enough tracked evidence before the agent asserts it.
Parameters* required
claimstringThe claim text to verify
check_operational_integrityEvaluate whether the current repo state is safe for PR, merge, release, and publish operations.5 params
Evaluate whether the current repo state is safe for PR, merge, release, and publish operations.
Parameters* required
baseBranchstringProtected base branch to compare against (defaults to main)
commandstringOptional git, PR, or publish command to evaluate against the current governance state
repoPathstringOptional repository path to inspect
requirePrForReleaseSensitivebooleanWhen true, release-sensitive changes on non-base branches require an open PR
requireVersionNotBehindBasebooleanWhen true, release-sensitive changes cannot lag behind the base branch package version
workflow_sentinelPredict pre-action workflow risk, blast radius, and remediations before a tool call executes.8 params
Predict pre-action workflow risk, blast radius, and remediations before a tool call executes.
Parameters* required
baseBranchstringOptional protected base branch override (defaults to main)
changedFilesarrayOptional affected-file list used to estimate blast radius
commandstringOptional shell command when toolName is Bash
filePathstringOptional primary file path for edit-like tools
repoPathstringOptional repository path used for git-aware integrity checks
requirePrForReleaseSensitivebooleanWhen true, release-sensitive changes on non-base branches require an open PR
requireVersionNotBehindBasebooleanWhen true, release-sensitive changes cannot lag behind the base branch package version
toolNamestringTool being assessed, such as Bash, Edit, or Write
register_claim_gateRegister a custom claim verification rule in local runtime state without editing tracked repo config.3 params
Register a custom claim verification rule in local runtime state without editing tracked repo config.
Parameters* required
claimPatternstringRegex pattern that should trigger claim verification
messagestringCustom message returned when evidence is missing
requiredActionsarrayTracked actions that must be present before the claim is verified
gate_statsGet gate enforcement statistics -- blocked count, warned count, top gates
Get gate enforcement statistics -- blocked count, warned count, top gates
No parameter schema in public metadata yet.
dashboardGet full ThumbGate dashboard -- Harness Score, gate stats, prevention impact, proof, and system health
Get full ThumbGate dashboard -- Harness Score, gate stats, prevention impact, proof, and system health
No parameter schema in public metadata yet.
org_dashboardOrg-wide multi-agent dashboard — shows all active agents, gate decisions, adherence rates, risk agents, and top blocked gates across the organization. Team rollout: full visibility. Free preview: limited to 3 agents.1 params
Org-wide multi-agent dashboard — shows all active agents, gate decisions, adherence rates, risk agents, and top blocked gates across the organization. Team rollout: full visibility. Free preview: limited to 3 agents.
Parameters* required
windowHoursnumberLookback window in hours (default 24)
settings_statusResolve managed, user, project, and local ThumbGate settings with per-field origin metadata for policy visibility.
Resolve managed, user, project, and local ThumbGate settings with per-field origin metadata for policy visibility.
No parameter schema in public metadata yet.
commerce_recallRecall past feedback filtered by commerce categories (product_recommendation, brand_compliance, sizing, pricing, regulatory). Returns quality scores alongside memories for agentic commerce agents.3 params
Recall past feedback filtered by commerce categories (product_recommendation, brand_compliance, sizing, pricing, regulatory). Returns quality scores alongside memories for agentic commerce agents.
Parameters* required
categoriesarrayCommerce categories to filter (default: all commerce categories)
limitnumberMax memories to return (default 5)
querystringProduct or brand context to find relevant past feedback
get_business_metricsRetrieve high-level business metrics (Revenue, Conversion, Customers) from the Semantic Layer.1 params
Retrieve high-level business metrics (Revenue, Conversion, Customers) from the Semantic Layer.
Parameters* required
windowstringAnalytics window (today, 7d, 30d, all)
describe_semantic_entityGet the canonical definition and state of a business entity (Customer, Revenue, Funnel).1 params
Get the canonical definition and state of a business entity (Customer, Revenue, Funnel).
Parameters* required
typestringone of
Customer · Revenue · Funnelestimate_uncertaintyEstimate Bayesian uncertainty for a set of tags based on past feedback.1 params
Estimate Bayesian uncertainty for a set of tags based on past feedback.
Parameters* required
tagsarrayTags to analyze for uncertainty
session_handoffWrite a session handoff primer that auto-captures git state (branch, last 5 commits, modified files), last completed task, next step, and blockers. The next session reads this automatically for seamless context continuity.6 params
Write a session handoff primer that auto-captures git state (branch, last 5 commits, modified files), last completed task, next step, and blockers. The next session reads this automatically for seamless context continuity.
Parameters* required
blockersarrayOpen blockers or unresolved issues
customContextstringAny additional context for the next session
lastTaskstringWhat was completed this session
nextStepstringExact next action for the next session
openFilesarrayKey files being worked on
projectstringProject name (auto-detected from cwd if omitted)
session_primerRead the most recent session handoff primer to restore context from the previous session. Call at session start.
Read the most recent session handoff primer to restore context from the previous session. Call at session start.
No parameter schema in public metadata yet.
list_harnessesList natural-language harness specs for portable workflow control, proof-backed verification, and GTM execution.1 params
List natural-language harness specs for portable workflow control, proof-backed verification, and GTM execution.
Parameters* required
tagstringOptional tag filter such as verification, acquisition, or workflow.
run_harnessExecute a natural-language harness through the async job runner with checkpoints, verification, and proof-backed outcomes.3 params
Execute a natural-language harness through the async job runner with checkpoints, verification, and proof-backed outcomes.
Parameters* required
harnessstringHarness id or file basename to execute.
inputsobjectOptional input overrides for template variables.
jobIdstringOptional stable job id for the resulting runtime.
scheduleCreate, list, or delete scheduled tasks. Supports natural language scheduling like "daily 9:00", "weekly monday 8:30", "hourly". Installs as macOS LaunchAgent or Linux crontab.6 params
Create, list, or delete scheduled tasks. Supports natural language scheduling like "daily 9:00", "weekly monday 8:30", "hourly". Installs as macOS LaunchAgent or Linux crontab.
Parameters* required
actionstringSchedule actionone of
create · list · deletecommandstringNode.js code to execute on schedule
descriptionstringWhat this schedule does
namestringSchedule name/ID
schedulestringSchedule spec: "daily 9:00", "weekly monday 8:30", "hourly"
workingDirectorystringWorking directory for the command
user_profileManage persistent user profile — preferences, style, domain knowledge that persists across sessions. Actions: add, remove, replace, view.3 params
Manage persistent user profile — preferences, style, domain knowledge that persists across sessions. Actions: add, remove, replace, view.
Parameters* required
actionstringProfile actionone of
add · remove · replace · viewcontentstringContent to add or new content for replace
old_textstringSubstring to match for remove/replace
session_searchSearch past session notes and conversations using full-text search. Returns relevant sessions from the SQLite FTS5 index for cross-session recall.2 params
Search past session notes and conversations using full-text search. Returns relevant sessions from the SQLite FTS5 index for cross-session recall.
Parameters* required
limitnumberMax results to return (default 10)
querystringSearch query to find relevant past sessions
open_feedback_sessionOpen a feedback session after thumbs up/down. Follow-up messages will be captured for 60s.3 params
Open a feedback session after thumbs up/down. Follow-up messages will be captured for 60s.
Parameters* required
feedbackEventIdstringThe feedback event ID from capture_feedback
initialContextstringsignalstringone of
up · downappend_feedback_contextAppend a follow-up message to an open feedback session. Call this when the user types additional context after giving thumbs up/down.3 params
Append a follow-up message to an open feedback session. Call this when the user types additional context after giving thumbs up/down.
Parameters* required
messagestringThe follow-up message from the user
rolestringone of
user · assistantdefault: usersessionIdstringfinalize_feedback_sessionFinalize a feedback session and re-infer the lesson with all follow-up context.1 params
Finalize a feedback session and re-infer the lesson with all follow-up context.
Parameters* required
sessionIdstringwebhook_deliverSend a message to Teams, Slack, or Discord via webhook. Use for status reports, alerts, and notifications.4 params
Send a message to Teams, Slack, or Discord via webhook. Use for status reports, alerts, and notifications.
Parameters* required
messagestringMessage body (markdown supported)
platformstringTarget platformone of
teams · slack · discordtitlestringMessage title
webhook_urlstringWebhook URL for the target channel
reflect_on_feedbackRun a post-mortem analysis on negative feedback. Returns a proposed rule and recurrence info.4 params
Run a post-mortem analysis on negative feedback. Returns a proposed rule and recurrence info.
Parameters* required
contextstringOne-line context from the caller
conversationWindowarrayLast 5-10 conversation turns before the feedback signal.
feedbackEventIdstringID of a previously captured feedback event
whatWentWrongstringWhat the caller said went wrong
report_product_issueReport a bug, suggestion, or complaint about ThumbGate itself (not project feedback). Auto-files a GitHub issue with system context. Use when the user expresses frustration or requests a feature for the thumbgate tool.3 params
Report a bug, suggestion, or complaint about ThumbGate itself (not project feedback). Auto-files a GitHub issue with system context. Use when the user expresses frustration or requests a feature for the thumbgate tool.
Parameters* required
bodystringDescription of the problem or suggestion, in the user own words
categorystringIssue categoryone of
bug · feature · questiontitlestringShort issue title (e.g. "Gate blocks valid migration")
run_managed_lesson_agentRun the LLM-powered lesson inference and rule generation agent over accumulated feedback. Requires ANTHROPIC_API_KEY for LLM mode; falls back to heuristics if unavailable.3 params
Run the LLM-powered lesson inference and rule generation agent over accumulated feedback. Requires ANTHROPIC_API_KEY for LLM mode; falls back to heuristics if unavailable.
Parameters* required
dryRunbooleanPreview what would be written without persisting
limitnumberMax feedback entries to process (default: 20)
modelstringOverride the Claude model (default: claude-haiku-4-5)
managed_agent_statusShow status of the last managed lesson agent run: entries processed, lessons created, gates promoted, and total runs.
Show status of the last managed lesson agent run: entries processed, lessons created, gates promoted, and total runs.
No parameter schema in public metadata yet.
run_self_distillRun the self-distillation agent to auto-evaluate recent agent sessions and generate improvement lessons without human feedback. Reads conversation logs, detects success/failure signals, and persists lessons.3 params
Run the self-distillation agent to auto-evaluate recent agent sessions and generate improvement lessons without human feedback. Reads conversation logs, detects success/failure signals, and persists lessons.
Parameters* required
dryRunbooleanIf true, analyzes but does not persist lessons
limitnumberMax conversation logs to process (default 20)
modelstringLLM model to use for analysis (requires ANTHROPIC_API_KEY)
self_distill_statusShow status of the last self-distillation run: sessions analyzed, lessons generated, signals detected.
Show status of the last self-distillation run: sessions analyzed, lessons generated, signals detected.
No parameter schema in public metadata yet.
context_stuff_lessonsDump ALL prevention lessons into a single text block for context-window injection. Bypasses RAG/search — returns every lesson sorted by confidence. For most projects (20-200 lessons), fits in 1K-10K tokens.3 params
Dump ALL prevention lessons into a single text block for context-window injection. Bypasses RAG/search — returns every lesson sorted by confidence. For most projects (20-200 lessons), fits in 1K-10K tokens.
Parameters* required
formatstringOutput format (default: compact)one of
compact · fullmaxTokenBudgetnumberApproximate token budget (default: 10000)
signalstringFilter by signal typeone of
positive · negativeThumbGate 👍 👎
Self-improving pre-action firewall for AI coding agents
AI coding agents repeat mistakes — and one wrong tool call can wipe a directory, leak a key, or push broken code.
What it does
ThumbGate is the local-first Pre-Action Checks engine for AI coding agents. It runs in the PreToolUse hook to evaluate the proposed tool call before execution — so costly mistakes can be caught before they happen.
Tech memes (shareable)
Lightweight visuals for how agents fail without a pre-action gate:
| Meme | Meaning |
|---|---|
 | Unchecked tool calls ship destructive commands. |
 | A prompt is advice; a PreToolUse hook is enforcement. |
It hard-blocks detected secret leaks and two direct self-disable command classes by default — commands that terminate the ThumbGate gate process or enable its bypass environment override. Other high-risk classes (rm -rf, force-push, fetch-and-run, direct guardrail edits) warn and log by default. Set THUMBGATE_STRICT_ENFORCEMENT=1 for strict enforcement (warnings become hard denies).
| Verdict | Default behavior |
|---|---|
| ⛔ Hard-block | Detected secret leaks; process-kill/environment-override self-disable |
| 👎 Warn + log | rm -rf, git push --force, fetch-and-run, direct guardrail edits — warn by default |
| 👍 Allow | Everything else |
Accepted feedback is stored as local lessons. Repeated concrete failures can become prevention rules that promote from warnings to blocking gates. The firewall improves from operations without retraining the model. Prompt evaluation (npx thumbgate eval) turns accepted feedback into reusable eval cases and local proof reports.
Honest disclaimer: ThumbGate does not update model weights. It intercepts tool calls at runtime. Local-first — no cloud required for the enforcement path.
Works with Claude Code, Cursor, Codex, Gemini CLI, Amp, Cline, OpenCode, and other MCP agents.
Agent tries: rm -rf tests/
ThumbGate: 👎 WARN + LOG — "Never delete test directories"
Pattern matched: rm.*-rf.*tests
Source: your thumbs-down from last Tuesday
Strict mode: ⛔ DENY before tool execution
Agentic development cycle fit
Agentic development is becoming a loop: Guide → Generate → Verify → Solve. ThumbGate is the pre-action gate / pre-action boundary between generated intent and executed action.
Quick Start
Want a phased walkthrough with a verify step at every stage? Follow the Progressive Setup Guide.
Progressive wiring — prove the pipe before you turn matching on. Empty dashboard is success.
npx thumbgate init # Phase 1: hooks only
npx thumbgate doctor # verify: exits 0 only when PreToolUse hook is wired (hidden metric = hook install, not gate count)
npx thumbgate dashboard --open # Phase 2: open local HTML; empty stats are OK
npx thumbgate capture --feedback=down --context="Never run DROP on production tables" --what-went-wrong="agent proposed DROP" --what-to-change="require review for DROP"
Later DROP attempts in the same scope surface the check:
⚠️ Check fired: "Never run DROP on production tables"
Pattern: DROP.*production
Verdict: 👎 WARN + LOG (⛔ BLOCK when THUMBGATE_STRICT_ENFORCEMENT=1)
Numbered configs: config/progressive/. Guide: progressive wiring.
MCP / Glama / registry install (stdio)
Directories and clients that install ThumbGate as an MCP server must start stdio MCP, not the HTTP API:
npx -y thumbgate serve
- Equivalent:
npx -y thumbgate mcp - Do not use
npm startfor MCP — that launches the hosted HTTP API (src/api/server.js), not the agent-facing stdio server.
▶ 90-second demo · GIF walkthrough
Install for your agent
| Agent | Command | Enforcement |
|---|---|---|
| Claude Code | npx thumbgate init --agent claude-code | 🛡️ Hard — PreToolUse |
| Codex | npx thumbgate init --agent codex | 🛡️ Hard — pre_tool_use |
| Gemini CLI | npx thumbgate init --agent gemini | 🛡️ Hard — PreToolUse |
| ForgeCode | npx thumbgate init --agent forge | 🛡️ Hard — pre_tool_use |
| Cursor | npx thumbgate init --agent cursor | 💬 Advisory — MCP gate_check |
| Cline | npx thumbgate init --agent cline | 💬 Advisory — MCP + .clinerules |
| OpenCode | npx thumbgate init --agent opencode | 💬 Advisory — MCP gate_check |
| Any MCP agent | npx thumbgate serve | 💬 Advisory — MCP gate_check |
| Amp | npx thumbgate init --agent amp | 📝 Feedback capture |
Per-agent guides: Claude/Codex bridge · Codex profile · Cursor · MCP setup
Install scope: machine-wide vs per-project
| Scope | Command | Settings | Lessons | Best for |
|---|---|---|---|---|
| Machine-wide (default) | npx thumbgate init | ~/.claude/settings.json | ~/.claude/memory/feedback/ | Solo operators — same machine-local feedback store across repos |
| Per-project | npx thumbgate init --project | <repo>/.claude/settings.json | <repo>/.claude/memory/feedback/ | Client / compliance — separate dashboard / isolated lessons per repo |
Both scopes write mcpServers.thumbgate plus PreToolUse / UserPromptSubmit / PostToolUse / SessionStart hooks. Machine-wide is the right default for most developers. Cross-repo blocking is not automatic: a lesson learned in one project only applies elsewhere when you share the store (machine-wide) or export/import lessons.
MCP tools (surface): gate_check (read/evaluate proposed tool call), feedback capture + session tools (write), dashboard/stats (read). Destructive agent actions stay blocked/warned by PreToolUse — ThumbGate does not execute user shell commands for you.
Discoverable slash-commands — the guardrail layer for spec-driven agents
Spec-driven agent frameworks like GSD (get-shit-done) and GitHub Spec Kit plan and generate work. ThumbGate is the guardrail layer for spec-driven agents: it sits after the plan, on the boundary between a generated tool call and its execution — alongside GSD / Spec-Kit, not instead of them.
npx thumbgate init installs these into your agent palette:
| Command | What it does |
|---|---|
/thumbgate-dashboard | Open local project dashboard |
/thumbgate-guard | Turn last mistake into a hard prevention rule |
/thumbgate-rules | List active rules & lessons |
/thumbgate-blocked | Gate stats + enforcement matrix |
/thumbgate-protect | Branch governance + scoped approval |
/thumbgate-doctor | Health-check hooks, MCP, readiness |
Pricing & buyer paths
Free tier: 2 feedback captures/day (10 total) and up to 3 active auto-promoted prevention rules. Pro ($19/mo or $149/yr) is the individual tier for unlimited rules, history-aware lessons, linked feedback session flow, personal dashboard, and DPO export. Enterprise is custom and scoped after intake; hosted team lesson sync and a hosted org dashboard are not general availability.
| Free | Pro ($19/mo or $149/yr) | Enterprise | |
|---|---|---|---|
| Local CLI + PreToolUse | ✅ | ✅ | Scoped after intake |
| Feedback captures | 2 feedback captures/day (10 total) | Unlimited | Scoped after intake |
| Active auto-promoted rules | up to 3 active auto-promoted prevention rules | Unlimited | Scoped after intake |
| Personal dashboard + DPO export | — | ✅ | Reviewed during intake |
| Hosted team lesson sync | — | — | Not general availability |
| Hosted org dashboard | — | — | Not general availability |
Enterprise intake path: the Workflow Hardening Sprint scopes one repeated failure before any broader rollout commitment. Start intake →
Local technical path: install the CLI and use init plus the documented setup so Pre-Action Checks evaluate tool calls where the agent actually runs.
First-dollar activation path: open the ThumbGate GPT, paste the risky action, capture typed feedback (thumbs down: / thumbs up:). Native ChatGPT rating buttons are not the ThumbGate capture path. Ask: what repeated AI mistake would be worth catching before the tool executes?
Paid path for individual operators: ThumbGate Pro is the self-serve side lane for a personal dashboard and export-ready evidence.
Start free · Pro $19/mo · Live Dashboard · Team Sprint intake · Workflow Hardening Sprint · First Dollar Playbook
Popular buyer questions: AI search topical presence · Relational knowledge and AI recommendations · AI Mode ads for agent governance · MCP tool governance · AI agent pre-action approval gates · Background agent governance · GPT-5.5 model evaluation · Stop repeated AI agent mistakes · Browser automation safety · Native messaging host security · Autoresearch agent safety · Cursor guardrails · Codex CLI guardrails · Gemini CLI memory + enforcement · Google Cloud MCP guardrails · Roo Code alternative: migrate to Cline
How it works (short)
- Capture 👍/👎 feedback (CLI, MCP, linked feedback session flow /
open_feedback_session, or ThumbGate GPT) - Promote concrete lessons via history-aware lesson distillation into prevention rules
- Evaluate the next proposed tool call against active rules (literal/AST + local vectors)
- Allow / warn / deny before the tool runs
npx thumbgate brain --write # → .thumbgate/BRAIN.md (lessons + gates in one artifact)
Pro operators can invoke search_lessons through MCP and use npx thumbgate lessons from the CLI. History-aware feedback sessions and lesson search are Pro capabilities; Free does not include recall or search.
Architecture diagram & stack
flowchart LR
A["Agent tool call"] --> B{"Rule match?"}
B -- exact --> D["On-device gate"]
B -- semantic --> C["Local LanceDB"]
C --> D
D -- secret/kill --> E["⛔ Hard-block"]
D -- known-bad --> G["👎 Warn + log"]
D -- safe --> F["👍 Allow"]
Built-in checks
⛔ secret-exfiltration → hard-block (default)
⛔ self-protect-kill → hard-block (default)
⛔ self-protect-env → hard-block (default)
⚠️ force-push → warn; hard-block under strict
⚠️ protected-branch → warn; hard-block under strict
⚠️ unresolved-threads → warn; hard-block under strict
⚠️ package-lock-reset → warn; hard-block under strict
CLI cheatsheet
npx thumbgate init
npx thumbgate doctor
npx thumbgate capture up|down "<text>"
npx thumbgate lessons
npx thumbgate brain --write
npx thumbgate dashboard --open
npx thumbgate break-glass --reason="ThumbGate over-fired" # 5-min recovery
Pro: lesson + DPO export
# Portable lessons
curl -X POST http://localhost:3456/v1/lessons/export \
-H "Authorization: Bearer $THUMBGATE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"outputPath": "./lessons-export.json"}'
# DPO pairs for fine-tuning
curl -X POST http://localhost:3456/v1/dpo/export \
-H "Authorization: Bearer $THUMBGATE_API_KEY" \
-o dpo-pairs.jsonl
Tech Stack
| Layer | Tech |
|---|---|
| Runtime | Node.js ≥18 |
| Interfaces | MCP stdio, HTTP API, CLI |
| Storage | SQLite + FTS5, LanceDB vectors, JSONL logs |
| Intelligence | MemAlign dual recall, Thompson Sampling, local embeddings |
| Billing / host | Stripe, Railway |
| Execution | Railway, Cloudflare Workers, Docker Sandboxes |
| Governance | Workflow Sentinel, control plane, Docker Sandboxes |
Every Changeset is tied to the exact main merge commit and generates Verification Evidence for Release Confidence.
Integrations (compact)
| Surface | Start here |
|---|---|
| Open ThumbGate GPT | thumbgate.ai/go/gpt — ThumbGate GPT: start here. Paste agent actions, get advice + checkpointing. No, users do not have to keep chatting inside the ThumbGate GPT to use ThumbGate — the hard enforcement layer still runs where the work happens. |
| Install Codex Plugin | Open the Codex plugin install page: thumbgate.ai/codex-plugin · zip: thumbgate-codex-plugin.zip · plugins/codex-profile/INSTALL.md |
Claude Desktop .mcpb | latest release |
| VS Code / Open VSX | plugins/vscode-extension/README.md |
| Antigravity-compatible | plugins/antigravity-extension/INSTALL.md |
| JetBrains | plugins/jetbrains-plugin/README.md · JetBrains Marketplace path for the same runtime |
| ChatGPT App / GPT Action | thumbgate.ai/chatgpt-app |
| ThumbGate-Core (staging) | https://github.com/IgorGanapolsky/ThumbGate-Core — pre-release staging + a few internal cache scripts; not the product moat |
Docs
Full index: docs/INDEX.md
| Need | Link |
|---|---|
| Agent workflow contract | WORKFLOW.md |
| Ready-for-agent intake | .github/ISSUE_TEMPLATE/ready-for-agent.yml |
| Verification Evidence | docs/VERIFICATION_EVIDENCE.md |
| Release Confidence | docs/RELEASE_CONFIDENCE.md |
| Changeset strategy | docs/CHANGESET_STRATEGY.md |
| First Dollar Playbook | docs/FIRST_DOLLAR_PLAYBOOK.md |
| Security policy | SECURITY.md |
| Threat model | THREAT_MODEL.md |
| Federal / regulated | docs/FEDERAL.md |
| Commercial Truth | docs/COMMERCIAL_TRUTH.md |
| Issues / PRs | GitHub Issues · PR template |
FAQ (one-liners): Not a fine-tuner (runtime intercept only). Different from CLAUDE.md / .cursorrules (those are context; ThumbGate is an external allow/warn/deny before tools run).
Who builds this
Igor Ganapolsky — payments (Stripe/Connect), AI agent guardrails/MCP, Android + backends. Small number of contract slots: $120–150/hr, 1099, remote US. LinkedIn · thumbgate.ai
License
MIT — see LICENSE. Project policy: SECURITY.md · THREAT_MODEL.md.
Featured
Plug Mailtrap into your AI workflow and let it handle the email.
Connect Mailtrap MCP →Agent, run crypto. Access onchain data & trade routes via 1inch.
Install now →Make coding agent sessions - Searchable, Shareable, Vendor-neutral & Scored.
Try For Free →Your agent targets a perfect 10 Code Health score. Deterministic. Every commit.
Try For Free →Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →belt cli automatically finds the best tools and skills for your agent. image, video, music, tts...
one prompt install →