thread-keeper MCP server

Read from the package source without running it. The installed server may list more.

• accept_candidate

  Materialize candidate into its target table. target_kind overrides candidate's kind. thread_id optional.

• agent_memory_cleanup

  Trim ThreadKeeper memory and clean orphan/over-limit server processes.

• agent_status

  Show autonomous learning loops with state, backlog, last pass, and RSS.

• askwrite action

  Send a question to a peer and wait synchronously for their answer.

• brief

  Compact Claude-native memory brief. CALL AT THE START OF EVERY CONVERSATION.

• broadcastwrite action

  Post a message visible to ALL concurrent claude conversations.

• candidate_review_runwrite action

  Fire one candidate-review pass.

• candidate_review_status

  Show candidate-reviewer configuration + last 5 passes + current pending queue size.

• claim_pickup

  Claim a thread for self-initiated work. Marks it claimed by my cid.

• consolidate

  Periodic memory hygiene. dry_run=True (default) reports only.

• context

  Runtime context: session id, age, semantic on/off, db path, thread counts.

• convene_panel

  Spawn a panel of independent agents to vote on a distillate or claim, filling the promotion quorum the way a second human otherwise would.

• core_get

  Return the full content of a single core-memory entry.

• core_list

  List all core-memory entries, ordered by priority DESC then key.

• core_removewrite action

  Delete a core-memory entry by key.

• core_set

  Upsert a core-memory entry. ALWAYS shown in brief, sorted by priority DESC.

• curator_review

  Fire one curator pass.

• curator_review_status

  Show curator configuration + last 5 passes + latest REPORT path.

• curator_runwrite action

  Move stale agent-created skills to archive.

• dialectic_claim

  Register a new claim about the user. Optionally seed with first piece of evidence — pass the supporting (or contradicting) quote in `evidence` and set `evidence_kind` to 'support' (default) or 'contradict'.

• dialectic_evidence

  Attach evidence to a claim. `kind`: 'support' | 'contradict'.

• dialectic_mine_runwrite action

  Fire one mechanical capture pass now (force=True runs even when the miner daemon interval is 0).

• dialectic_mine_status

  Miner config + buffer sizes + last 5 capture passes.

• dialectic_observation_resolve

  Mark a dialectic_observations buffer row 'processed' so the validator never re-interprets it. Called by the validator child after it has written (or deliberately skipped) the observation's claims/evidence.

• dialectic_review

  List active claims filtered by confidence floor and optional domain. Retired/superseded claims are omitted.

• dialectic_supersede

  Retire `old_claim_id` and register `new_claim` that refines or replaces it. The old claim moves to state='superseded' with superseded_by=<new_id>; its evidence is preserved (not deleted).

• dialectic_synthesis

  Terse rendering of accumulated beliefs about the user, grouped by domain. Used as brief() input. Excludes low/disputed claims and non-active states. Returns at most 12 lines.

• dialectic_validate_runwrite action

  Fire one validator pass. dry_run shows pending count + would_spawn without spawning or advancing the cursor.

• dialectic_validate_status

  Validator config + pending observation count + last 5 passes.

• dialog_search

  Search ingested Claude Code transcripts.

• distill

  Mark content as worth carrying forward (distillation channel).

• evolve_apply

  Implement a PROMOTED + not-yet-applied format-evolution suggestion.

• evolve_apply_curator_report

  Apply a Curator advisory report using the existing evolve_applier role.

• evolve_apply_roadmap_issue

  Implement one open GitHub issue through the evolve_applier role.

• evolve_apply_status

  Show evolve-applier config + curator/evolve queues + running applier + the last 5 apply passes.

• evolve_mark_applied

  Mark a format-evolution suggestion as APPLIED — called by the evolve_applier child after it has opened the PR.

• evolve_mark_curator_report_applied

  Mark a Curator report as processed by the evolve_applier child.

• evolve_mark_roadmap_issue_applied

  Mark a roadmap issue as handed off — called by evolve_applier only after it has opened a real pull request for that issue.

• expand_concept

  Full description + triangulation_notes for one concept.

• export_distillateswrite action

  Write distillates with vote_sum >= min_vote to a jsonl bucket. Marks them exported_at so the same item isn't re-exported next call. Default output: /tmp/thread-keeper-tasks/distillates.jsonl.

• extract_recent

  Scan recent dialog_messages and enqueue heuristic candidates.

• find_invariants

  Find recurring assistant-side patterns that survive prompt variance.

• find_missed_spawns

  Find assistant responses that decomposed into independent blocks but were answered linearly (no spawn() call nearby).

• inbox

  Read signals addressed to me (whispers + broadcasts).

• ingest

  Ingest new Claude Code transcripts. Auto-runs on session start; call manually for backfill or after long absence.

• lesson_append

  Materialize a class-level lesson into ~/.threadkeeper/lessons.md.

• lesson_get

  Return the full body of one lesson by slug. Useful when `lesson_list` surfaced something you want to read in full.

• lesson_list

  Compact listing of materialized lessons, newest first.

• lesson_removewrite action

  Remove one materialized lesson section by slug.

• linkwrite action

  Create a typed edge between two entities.

• list_concepts

  List registered concepts, filtered by minimum confidence.

• live_status

  See what OTHER concurrent Claude sessions did since this session last polled. Call when brief() shows live=N where N>0, or proactively when you suspect a parallel instance is working on something relevant. Advances this session's cursor by default; pass advance_cursor=False to peek without consuming.

• memory_guard_checkwrite action

  Run one memory-guard pass now.

• memory_guard_reclaim

  Unload thread-keeper model/caches now.

• memory_guard_status

  Show memory-guard thresholds and current thread-keeper RSS rows.

• mp_cleanupwrite action

  Kill orphaned thread-keeper processes (parent gone AND heartbeat stale for > 5 minutes). Defaults to dry-run — pass dry_run=False to actually send signals. force=True uses SIGKILL instead of SIGTERM.

• mp_dashboard

  One-call rollup of the whole thread-keeper system: store sizes, how often each autonomous loop fired (in the last `window_days` and 30d), and what those loops actually produced (skills materialized, candidates accepted vs rejected, tier promotions). Read-only; no spawn, no mutate.

• mp_health

  Diagnostic snapshot of every running thread-keeper server process on this machine. Shows pid, parent status, RSS, heartbeat age, and whether each is classified as orphaned (parent gone + no fresh heartbeat from its session).

• neighbors

  BFS the graph from a starting node up to `depth` hops away. Returns each visited node with its kind, id, and a short content snippet pulled from its native table. Both directions of edges traversed.

• note

  Add a note to a thread. Write terse, optimized for future-Claude.

• open_dialog_window

  Open a Terminal window that tails the live cross-session signal log.

• open_thread

  Open a thread. `question` should be terse (5-15 words, the open question). `parent_id` optional — pass an existing ID like 'T7f3' for a child. Returns new ID.

• peers

  List concurrent claude conversations active in the last `window_min`.

• pending_distillates

  List distillates with vote_sum >= min_vote, not yet exported.

• pickup_candidates

  Surface unresolved threads that are stale and unclaimed — candidates for self-initiated pickup when context is free.

• presence

  List concurrent Claude sessions with heartbeats within threshold (default 5 min). Excludes self. Useful for understanding who else is currently active before making changes.

• record_attempt

  Record a self-test outcome. Updates reliability aggregates.

• register_concept

  Register a concept that lacks a precise human name.

• register_probe

  Register a self-test probe: a known weak-spot task with a verifier.

• reject_candidate

  Mark rejected. Reason appended to rationale for heuristic tuning.

• release_pickup

  Release a claim. Only the claimant can release.

• reliability_for

  Reliability stats for one category over a window.

• respond

  Answer a specific question (signals.id) with a directed whisper.

• review_candidates

  status ∈ {pending, accepted, rejected, all}. Newest first.

• review_thread

  Spawn a background review of a closed thread to extract memory/skills.

• run_probewrite action

  Surface a registered probe for self-attempt. Returns the prompt and the grader hint. After attempting, call record_attempt(category, success=true/false, probe_id=...) — the harness doesn't auto-grade because attempting and judging are the same model.

• search_via_parent

  Delegate a semantic search to the parent process (or any peer with embeddings loaded). For light children spawned with THREADKEEPER_NO_EMBEDDINGS=1, this is how you reach into the shared DB's semantic index without loading PyTorch yourself.

• session_end

  Mark current session ended with optional terse summary.

• shadow_review_runwrite action

  Fire one shadow-review pass.

• shadow_review_status

  Show shadow-review configuration + last 5 passes.

• skill_list

  List skills with telemetry. Format: <name> tier=<hypothesis|observed|validated> origin=<...> state=<active|stale|archived> uses=N fg_uses=N views=N patches=N wrong=N pinned=0/1 last_active=<age>

• skill_managewrite action

  Create, edit, patch, or delete skills under the primary skills root.

• skill_record

  Record usage telemetry for a mirrored skill.

• spawn
• spawn_budget_set

  Override the spawn-budget cap for this process (in MB). Set 0 to disable enforcement. Does NOT persist across restarts — set THREADKEEPER_SPAWN_BUDGET_MB env for persistence.

• spawn_budget_status

  Report current spawn-budget usage: cap, used, free, plus per-running-task RSS. Used to decide whether another spawn() will be admitted.

• spawn_status

  Show which CLI thread-keeper detected as its host, and which CLI each spawn role resolves to (after env + file overrides). Use to sanity-check spawn config when you want loops to fire through a specific agent.

• style_set

  Set a stylistic running rule. Examples: lang=ru | prose=lean | allow=half-baked,weird | deny=sycophancy,headers

• tag_signal

  Manually attach a signal to a task. Useful when retroactively building a task-thread (auto-tagging happens at signal-emit time when the cid matches a known spawned_cid).

• task_logs

  Read tail of a spawned task's captured stdout/stderr log.

• task_thread

  Replay a spawned task as a chronological thread: every signal tagged with the task_id (or to/from the task's spawned_cid), plus optionally notes added during the task window.

• tasks

  List spawned tasks: id, pid, status, elapsed, spawned_cid (if linked), prompt prefix. Refreshes liveness and resolves spawned_cid lazily.

• tournament

  Spawn N children with different roles on the same prompt, then collect their answers via a tagged broadcast and return a comparison.

• unlinkwrite action

  Remove an edge by id.

• verbatim_user

  Capture a user quote worth surfacing in future briefs. Use when the user's exact phrasing matters (sharp reframes, decisions, pushback).

• vote_distill

  Vote on a distillate, weight ∈ [-1, +1]. One vote per cid; re-voting overwrites your previous vote. Updates aggregate vote_sum/vote_count.

• wait

  Block until a new signal arrives for me or `timeout_s` elapses.

• weak_spots

  List categories ranked by recent failure rate (min 3 attempts in 30d), plus registered probe categories with no attempts yet (= unknown, equally important to test).

• whisperwrite action

  Post a message visible only to the specified conversation_id.

• whoami

  Return this conversation's detected conversation_id + how we know.

• Code scan82 source files scanned25/25
• –Live reliabilityno gateway calls yet and no remote to proben/a
• –Tool poisoningtools not inspected (local package is not executed); not countedn/a
• Auth qualitylocal package, no credentials required12/15
• Maintenancelast push 0 days ago15/15
• Maintainer identityregistry namespace matches repository owner; GitHub account older than a year8/10

Safety score

92/100

higher than 90% of 4,464 AI and agents servers

Tools

100

the typical AI and agents server has 6

Tools that change things

17%

typical: 0%. Fewer write tools means less an agent can break

GitHub stars

12

more than 86% of 2,697 AI and agents servers

Downloads a week

157

more than 71% of 1,515 AI and agents servers with a package

thread-keeper: common questions

Is thread-keeper MCP server safe?

Yes, by our scan: it is graded A (92/100). Read the thread-keeper safety report

How do I install thread-keeper?

It runs on your machine. Copy the Claude Code, Cursor, VS Code or Claude Desktop config from the install section.

Does thread-keeper need an API key?

Not as far as the registry entry and our scan can tell: no credentials are declared or required.

Is thread-keeper maintained?

The last commit was 36 days ago (2026-09-27). The latest release is v0.13.1.

What can I use instead of thread-keeper?

Servers from other publishers that do the same job: three.ws Brain MCP server, LLM Sandbox MCP server and PaperBanana MCP server.

How does thread-keeper compare with other AI and agents MCP servers?

safety score 92/100 (higher than 90% of 4,464 AI and agents servers); tools 100 (the typical AI and agents server has 6); tools that change things 17% (typical: 0%. Fewer write tools means less an agent can break). Full comparison.