io.github.plur-ai/plur

Gateway to the 33 PLUR operations that are not top-level tools under the current profile (collapsed into one dispatch tool so Cursor's ~40-tool-per-workspace limit is not exhausted by PLUR alone). A plur_* name missing from tools/list means it moved HERE — not that the MCP is unavailable. Calling convention: { action: "<tool name>", args: { ...that tool's normal arguments } } — same arguments, same validation, same result as a direct call. Example: { action: "plur_recall_hybrid", args: { query: "deploy checklist" } }. Send { action: "help" } for every action's one-line description and argument schema. Actions, grouped — inject: plur_inject, plur_inject_hybrid · packs: plur_packs_discover, plur_packs_export, plur_packs_install, plur_packs_list, plur_packs_preview · stores: plur_stores_add, plur_stores_list · sync: plur_sync, plur_sync_status · other: plur_capture, plur_episode_to_engram, plur_extract_meta, plur_history, plur_ingest, plur_learn_batch, plur_meta_engrams, plur_outbox, plur_pin, plur_profile, plur_promote, plur_provenance, plur_recall_hybrid, plur_report_failure, plur_rescope, plur_scopes_discover, plur_session_scope, plur_similarity_search, plur_suggest_scope, plur_tensions, plur_timeline, plur_validate_meta.

Diagnose the PLUR ENGINE (embedder, hybrid search, remote-store auth) — not hook/MCP wiring. Reports whether the embedding model loaded, whether hybrid search is fully operational, and — for any configured enterprise/remote store — whether its auth is valid (probes /api/v1/me and decodes token expiry), so a dead or soon-to-expire token surfaces instead of hiding behind a "healthy" report. Run this first when recall feels off or team engrams stop syncing. Does NOT check .cursor/mcp.json, .cursor/hooks.json, or the live MCP tool count — for that, run the `plur doctor` CLI command in a terminal (a different, more thorough check with the same name).

Rate an engram's usefulness — trains injection relevance over time. Supports single or batch mode.

Retire an engram by ID or search term — marks it as no longer active without deleting history

Create an engram — record a reusable learning, preference, or correction. A write is never suppressed by similarity: exact content-hash duplicates NOOP, and anything merely SIMILAR is written and reported back in `dedup.near_duplicates` (closest existing engrams, their cosine scores, and a preview of each neighbour's own statement — read them before moving on; that is what they are for) so you can supersede or merge deliberately. High similarity is a reason to look, not a decision — cosine cannot tell a duplicate from a correction of it. Multi-agent note: in an orchestration that spawns subagents, have the PARENT session own plur_learn writes — spawned subagents should return their findings as text for the parent to persist, rather than each calling plur_learn (tool availability is not guaranteed in every subagent context). See plur-ai/plur#281.

Uninstall an engram pack by name — removes the pack and all its engrams

Search engrams by topic. Default mode is hybrid (BM25 + local embeddings via RRF) — set mode:"keyword" for BM25-only. Local search plus, when a configured enterprise store is part of the current project/work, one live timeout-bounded recall per remote host merged in (a `remote_stores` block + warning appears when a host is degraded; no host configured or implicated = fully local). Note: a project-scope filter also returns personal-family engrams (local, global, user:*, agent:*); an explicit scope=global recall returns ALL personal-family engrams — wider than scope=global INJECT, which is targeted to the global namespace only.

Counted report of what your memory retrieved for you: engrams stored, how many were retrieved and how often, which are most relied on, and how much of the store is dormant. Local and read-only; every figure is directly counted, never estimated. IMPORTANT when relaying to the user: `activation_rate` is COVERAGE over the logging window (≈ how much of the store was surfaced), NOT a quality or effectiveness score — it is naturally low and FALLS as more engrams are added, so never present it as "memory is N% effective". A `summary` line is included; prefer relaying that. NOT plur_provenance: this counts usage across the whole store; plur_provenance says where a single memory came from.

End a session. BEFORE calling this tool, review the conversation and extract learnings: 1. Corrections the user made ("no, use X not Y") → type: behavioral 2. Preferences stated ("always X", "never Y") → type: behavioral 3. Codebase patterns discovered (naming, structure, conventions) → type: architectural 4. Technical facts learned (API quirks, config, gotchas) → type: procedural 5. Terminology defined or clarified → type: terminological Include at least one engram_suggestion if ANYTHING was learned. An empty suggestions array means nothing worth remembering happened — this should be rare.

Start a session — inject relevant engrams for your task. Call at the beginning of every session.

Return system health — running version, engram count, episode count, pack count, storage root. Optionally filter engram counts by domain prefix and/or creation date.

Purge all conflict relations from local engrams — removes accumulated false positives from the legacy tension-detection system