Wyrm MCP server

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

• buddy

  Use when another buddy-compatible MCP server (or the operator) asks Wyrm for a brief, data-grounded status reply - the Buddy Protocol v1.0 well-known entry point. Companion params switch to the full companion reply; bare calls stay protocol-pure. Cycle-protected.

• wyrm_activate

  Activate a license key to unlock Pro/Team/Enterprise features

• wyrm_agent_init

  Bootstrap Wyrm's autonomous agent loop in one call. Starts the wyrm-loop daemon as a detached background process so subsequent ticks happen automatically every interval_seconds. Optionally seeds a first goal in the same call. Idempotent: if already running, returns status without spawning a second instance.

• wyrm_agent_restart

  Stop the daemon and start it again with optional new settings. Useful when changing interval_seconds or max_steps without manually stopping first.

• wyrm_agent_status

  Inspect the autonomous agent: is the wyrm-loop daemon running, what was the last action, how many active goals + total iterations. Use to confirm bootstrap worked.

• wyrm_agent_stop

  Stop the autonomous agent daemon (SIGTERM with grace, then SIGKILL). Goals remain in DB and can be resumed by calling wyrm_agent_init again.

• wyrm_analytics_dashboard

  View usage analytics dashboard: tool calls, tokens, costs, cache efficiency (Pro+)

• wyrm_audit_export

  Export a date-ranged tamper-evident audit bundle (JSON). Includes full chain so an external auditor can verify offline without DB access. Use range_start='2026-01-01' for SOC2 audit windows.

• wyrm_audit_log

  Append a hash-chained audit event. Each entry references the previous entry's sha-256, so tampering breaks the chain. Optional Ed25519 signature layer. Used for SOC2/HIPAA-grade audit.

• wyrm_audit_verify

  Walk the entire audit_log chain and verify every prev_hash / row_hash. Returns ok=true if intact, or first_invalid_id when tampering is detected. Optional public_keys also verify Ed25519 signatures.

• wyrm_auto_capture

  Auto-extract durable memories from freeform text (a session transcript, notes, a chat) into the REVIEW QUEUE. A LOCAL extractor (a configurable Ollama model via WYRM_EXTRACT_MODEL or the `model` arg if set, else a deterministic heuristic) pulls out truths, failures, decisions, patterns, and lessons as needs_review candidates; approve/reject with wyrm_review (nothing auto-trusted). Closes the 'just

• wyrm_buddy

  Friendly, data-grounded coding companion. Returns a brief greeting + project state + suggested next step sourced entirely from Wyrm's real data. Every claim cites a real DB row ID; no hallucinated encouragement. Personas: wyrm (default, dragon-themed), drogo (gruff), berlin (precise), sumair (warm), custom. Moods: normal / celebratory / stuck / quiet (auto-detected). Sizes: full (10-line PhantomDr

• wyrm_call_external

  Use to call a tool on an outbound MCP server registered via wyrm_mcp - lazily spawned over stdio, result returned and logged to external_call_log.

• wyrm_capabilities

  Use to learn what your memory system can actually do - feature inventory, why each matters, runtime state, and which tool to reach for. Call at session start instead of re-discovering the surface.

• wyrm_capture

  Use to save anything worth keeping - the single write funnel. Auto-classifies notes, decisions, and lessons into quest, ground truth, or memory; risky writes queue for wyrm_review. mode=direct (store a proven pattern or lesson learned so future agents reuse it), extract (pull memories from a conversation transcript or freeform text), trace (harvest a session JSONL / tool-call log into run-tagged r

• wyrm_capture_trace

  Harvest a harness's working trace into run-tagged review-queue candidates (OFFLINE). Ingests a Claude Code session transcript (JSONL), a ~/.dragon/traces file/dir, or WYRM_TRACE_TOOL_CALLS event output. The SAME local extractor as auto-capture (Ollama WYRM_EXTRACT_MODEL if set, else deterministic; never a cloud LLM) pulls durable truths/failures/decisions/patterns/lessons; secrets (API keys, token

• wyrm_check_updatewrite action

  Check whether a newer wyrm-mcp is available on npm. Cached 24h by default; pass force:true to bypass cache.

• wyrm_cloud_backup

  Legacy whole-DB snapshot backup to encrypted cloud storage (Pro+): backup, restore, list, prune. This is a full-database snapshot (last-write-wins on restore), NOT the sovereign per-row sync; for multi-device memory use wyrm_cloud_sync action=sovereign. Keep this for simple point-in-time backups.

• wyrm_cloud_sync

  Cloud sync. The SOVEREIGN path is the promoted one: zero-knowledge, per-row AES-256-GCM E2E-encrypted delta sync with tombstone propagation; the cloud never sees plaintext. Use action 'sovereign' to sync (push+pull), 'sovereign-preview' for a dry run, 'sovereign-status' to show device + sync state. For hands-off periodic sync, 'sovereign-daemon-start' / 'sovereign-daemon-stop' / 'sovereign-daemon-

• wyrm_context_build

  Use to build a compact context brief for the current task under a token budget - relevant patterns, lessons, and truths in one formatted block; max_tokens elides low-score items to recallable stubs.

• wyrm_cost_report

  View estimated API cost report for current billing period (Pro+)

• wyrm_data_batch_insertwrite action

  Batch insert multiple data points efficiently

• wyrm_data_categories

  List all data categories for a project

• wyrm_data_insertwrite action

  Insert data into the data lake for a project

• wyrm_data_query

  Query data from the data lake

• wyrm_decided_because

  Use to link a choice or decision to the assumption or evidence it was based on - records a causal edge (truth/quest/artifact/session) so stale foundations invalidate downstream.

• wyrm_decision_downstream

  List all decisions that depend on the given source. Use before invalidating a truth to see the blast radius.

• wyrm_decision_invalidate

  Cascade-invalidate all decisions downstream of a source (recursive). Use when a foundational truth turns out to be wrong. Returns the set of affected (kind, id) pairs.

• wyrm_decision_upstream

  List all sources a decision was based on (incoming edges). Walks the 'why did we choose this?' chain.

• wyrm_digest

  Render a plain-English summary of what Wyrm has done for the operator over a period. Shows counts (sessions tracked, quests completed, repeated failures blocked, ground truths set, references clipped, hours logged) and highlights so a non-technical operator can see Wyrm's actual contribution. Use this when the user asks 'what did Wyrm do?', 'is Wyrm doing anything?', or to demonstrate value to ske

• wyrm_distill

  Distill a work session into candidate memory artifacts for review. Parses the session's objectives, decisions, and outcomes into structured artifacts (lessons, patterns, anti-patterns) and queues them for approval via wyrm_review. Use at session end to extract institutional knowledge.

• wyrm_embed

  Make Wyrm FIRST-PRIORITY, always-loaded memory across Claude Code and other CLIs, so it isn't ignored and is consulted proactively, not just when called. Prepends a managed 'Wyrm is your memory. Consult it FIRST.' directive to the TOP of each client's always-loaded instruction file (~/.claude/CLAUDE.md globally; project CLAUDE.md/AGENTS.md, and with allClients also .cursorrules/.windsurfrules/.git

• wyrm_encrypt_setup

  Set up or check encryption status for sensitive data (Pro+)

• wyrm_entity_add

  Add an entity to the knowledge graph (person, concept, tool, etc.)

• wyrm_entity_graph

  Get the neighborhood graph around an entity: all connected nodes within N hops

• wyrm_entity_linkwrite action

  Create a relationship between two entities in the knowledge graph

• wyrm_entity_merge

  Merge two entities: combines aliases, redirects relationships, deletes the source

• wyrm_entity_path

  Find the shortest path between two entities in the knowledge graph

• wyrm_entity_search

  Search for entities in the knowledge graph

• wyrm_events_publishwrite action

  Live Memory: manually publish an event onto a project's stream (e.g. a tool_call or presence marker). Failure-isolated: never throws. Payload is stored only when isShared is true (reference-only otherwise).

• wyrm_events_replicate

  Live Memory (Phase 3): one-shot cross-device replication with a peer Wyrm node: pulls the peer's shared events and pushes ours over the peer's HTTP surface. Privacy-gated (only is_shared events leave the box; reference-only rows carry no payload), with automatic dedup + echo-suppression. Watermarks persist, so repeat calls are incremental. Returns pulled/pushed/duplicate counts.

• wyrm_events_since

  Live Memory: pull events newer than a cursor for a project, in chronological order. Idempotent: safe to re-poll. Returns the events and the new head cursor to pass next time.

• wyrm_events_subscribe

  Use to START consuming a project's Live Memory event stream: returns the current head cursor plus a recent chronological window of events in one read-only call. Nothing persists server-side; poll wyrm_events_since with the returned cursor for new events. Returns disabled:true if WYRM_LIVE_MEMORY is off. Keywords: events, live memory, event stream, cursor, tail events, watch project, replication fe

• wyrm_failure_check

  Use before proposing or executing any fix, command, or approach - did this fail before? Is it a known landmine? Counter-pattern memory: returns unresolved matching failures (empty = safe to proceed); show the mistakes already made on a project with `wyrm failure list` (CLI). Subagents: check before retrying anything flagged.

• wyrm_failure_list

  List unresolved failures for a project (or globally), most-recent-first.

• wyrm_failure_record

  Use when an approach failed or something broke - log the dead end so nobody tries it again. Identical (scope, target, description) coalesces with occurrences++; when the root cause is fixed, stop flagging it with `wyrm failure resolve <id>` (CLI). Run-scoped failures quarantine until confirmed.

• wyrm_failure_resolve

  Mark a failure as resolved (the underlying issue was fixed). Keeps history but excludes from check() results.

• wyrm_feedback

  Record whether a recalled knowledge artifact was useful. Adjusts confidence over time: successful reuse boosts it, failure lowers it. Call after applying a recalled pattern or lesson.

• wyrm_goal_abandon

  Abandon a goal (won't be auto-pursued; record kept for audit).

• wyrm_goal_complete

  Manually mark a goal completed (e.g. when the agent reports done or the operator finishes it).

• wyrm_goal_iterations

  List the OODA iterations Wyrm has run for a goal: what it observed, decided, and resulted.

• wyrm_goal_list

  List goals, optionally filtered by status / project. Default: active goals, prioritized.

• wyrm_goal_pause

  Pause a goal; wyrm-loop will skip it until resumed.

• wyrm_goal_resume

  Resume a paused goal.

• wyrm_goal_set

  Set a persistent goal Wyrm will pursue across sessions. Goals carry title, success_criteria, deadline, priority, and a max_iterations cap. The wyrm-loop daemon picks the highest-priority active goal each tick.

• wyrm_harvest

  Auto-populate memory from artifacts you already produce. Walks a project (or ALL registered projects) and harvests durable facts from its docs (README/CLAUDE/AGENTS/ARCHITECTURE) plus recent commit subjects (git log) into the REVIEW QUEUE; approve/reject with wyrm_review, nothing is auto-trusted. Idempotent (re-runs skip what's already harvested). Use dryRun to preview. The fix for a thin corpus.

• wyrm_hours_report

  Hours worked, derived from sessions in a date range. Per-project breakdown + per-session line items. Falls back to a default-hour estimate for sessions with no content.

• wyrm_import_git

  Import a list of commits into the Wyrm memory review queue. Useful for onboarding a project's history or capturing recent work.

• wyrm_import_pr

  Import a pull request (title, body, review comments) into the Wyrm memory review queue to extract patterns and heuristics.

• wyrm_import_rules

  Import a .cursorrules / copilot-instructions file into Wyrm. Parses by section and creates ground truths (for constraint/rule language) or memory artifacts (for general guidelines).

• wyrm_inject_prompt

  Opt-in: inject a Wyrm session-prime instruction block into .github/copilot-instructions.md and/or .cursor/rules for a project. Uses managed markers so re-runs are safe.

• wyrm_intro

  Return a plain-English explanation of what Wyrm is and what it does. Aimed at HUMANS who don't yet understand Wyrm's value: co-founders, operators, designers, anyone who didn't read the docs. Call this when a human user asks 'what does Wyrm even do?' or seems unaware Wyrm is involved in their AI assistant's behavior. NOT for AI-agent self-orientation (use wyrm_capabilities for that).

• wyrm_invoice_generate

  Generate a markdown invoice from a date-ranged hours report. Per-session line items + per-project totals + grand total.

• wyrm_license

  View current license status, tier, and available features

• wyrm_list_projects

  List all registered projects with their status

• wyrm_maintenance

  Use for periodic upkeep of the memory store - archive old sessions, vacuum, prune stale events, sweep run quarantine. Admin-gated, destructive; fleets run it between waves.

• wyrm_mcp_disable

  Disable an MCP server config (kept in DB but won't be spawned).

• wyrm_mcp_list

  List registered MCP server configs.

• wyrm_mcp_register

  Register an outbound MCP server config. Wyrm can then call its tools via wyrm_call_external. e.g. server_name='github', command='npx', args=['-y', '@modelcontextprotocol/server-github'], env={GITHUB_PERSONAL_ACCESS_TOKEN: '...'}.

• wyrm_mcp_tools

  List the tools a registered server exposes (one-shot tools/list call over MCP).

• wyrm_migrate_prompt

  Rewrite the Wyrm-managed block in this project's AI-client instruction files (.cursor/rules, .github/copilot-instructions.md, CLAUDE.md, etc.) to the current canonical version. Operator-authored content outside the <!-- wyrm:start --> / <!-- wyrm:end --> markers is preserved verbatim. Dry-run by default; pass apply:true to write changes.

• wyrm_orchestrate_task

  Classify a task and get automatic orchestration recommendation (ensemble voting, parallel research, etc)

• wyrm_orchestration_config

  Get or update auto-orchestration configuration

• wyrm_orchestration_stats

  Get auto-orchestration effectiveness statistics and task distribution

• wyrm_presence_announce

  Use from every fleet agent to announce presence and heartbeat liveness: upserts a TTL row that orchestrators and sibling agents poll before assigning or stealing work, stamped with the ambient run_id and the agent's fleet role so the board reads per-run. Re-announce to heartbeat; dead sessions expire by TTL and are reaped by wyrm_maintenance. Announce BEFORE wyrm_quest_claim - claims require a pre

• wyrm_presence_list

  Use from an orchestrator to see which agents are live right now (heartbeat within TTL) before dispatching work - optionally scoped to a project or to agents touching a specific file, so file-level collisions surface before they happen.

• wyrm_presence_release

  Use on clean agent exit: removes the presence row and drops every quest claim the agent holds, so queued work is immediately re-claimable by the rest of the fleet instead of waiting out the TTL. Idempotent - safe to call twice.

• wyrm_project_context

  Get full context for a specific project including recent sessions, quests, and stored context

• wyrm_reference_add

  Clip a design reference (URL, image path, palette, or inline snippet) for inspiration. Stored per-project (or globally if no project path). Tag aggressively: tags drive future search and recall when building a new project in the same vein.

• wyrm_reference_list

  Browse saved design references. Filter by project or tag. Default: 50 most recent across all projects.

• wyrm_reference_search

  Full-text search across design references' title / notes / tags. Returns matching references sorted by recency.

• wyrm_reindex

  Rebuild embeddings for a project or all projects. Regenerates vectors for sessions, quests, data, and memory artifacts (the latter power hybrid recall).

• wyrm_remember

  Store a distilled knowledge artifact: proven patterns, lessons learned, anti-patterns, and reasoning traces. These are recalled automatically by wyrm_context_build to help AI models apply past knowledge to new tasks. Example: wyrm_remember({ projectPath: '/home/user/api', kind: 'anti_pattern', problem: 'Jest suite hung after adding the SSE server', validatedFix: 'close() the server in afterAll; it

• wyrm_replication

  Use to sync memory between machines and watch the live event mesh - action=start|stop|status|tick (daemon + peers; tokens via tokenEnv), conflicts (did the last sync leave conflicts to resolve?), resolve, publish, since, subscribe (watch a project's live activity feed for changes from other agents), replicate (one-shot: sync my memory over to another machine or peer), unshare (kind + id: make a re

• wyrm_scaffold_get

  Find the best reasoning scaffold for a task description. Returns the most relevant structured checklist, or null if no match meets the confidence threshold.

• wyrm_scaffold_save

  Save a reasoning scaffold: a structured checklist that guides step-by-step thinking for a specific problem type. Scaffolds are automatically surfaced by wyrm_context_build when the task matches, helping AI models apply proven reasoning patterns.

• wyrm_scan_projects

  Scan a directory for git projects and register them. Use this to auto-discover all projects.

• wyrm_self_updatewrite action

  Run `npm install -g wyrm-mcp@latest` to upgrade. Returns the install transcript. Requires the operator to have write access to the global npm prefix.

• wyrm_setup

  Auto-detect installed AI clients (VS Code, Claude Desktop, Cursor, Windsurf, Zed) and configure Wyrm's MCP server in all of them. Run this to connect Wyrm to a new AI or after switching providers.

• wyrm_spec_register

  Make spec-kit Wyrm-native. Reads a spec directory (spec.md title/summary + tasks.md task list) and creates one Wyrm quest per task, linked to a project, with a stored spec\u2192project link. Idempotent: re-running updates the same quests (deduped by a per-spec-task signature) instead of duplicating.

• wyrm_stats

  Use for a health snapshot of the memory database - row counts, vector index coverage, usage.

• wyrm_sync

  Sync database with .wyrm folders in all projects

• wyrm_token_deletewrite action

  Remove a design token from a project.

• wyrm_token_get

  Retrieve all design tokens for a project, optionally filtered by category. Returns the design system primitives the project is built on.

• wyrm_token_set

  Set a design token for a project. Tokens are colour / type / spacing / motion / shadow / radius / breakpoint / custom: the design system primitives the project is built on. Stored first-class so AI helpers working on creative tasks don't have to re-derive the system from scattered CSS each time. Upserts on (project_id, category, key).

• wyrm_tool_analytics

  Wyrm watches Wyrm: per-tool usage stats (volume, error rate, p50/p95 latency) over a configurable window. Use to find under-performing tools and tune retrieval/ranking. Pass mode='overview' for the total summary, 'usage' for per-tool, 'errors' for recent error sample.

• wyrm_truth_get

  Use to read the established facts and ground rules for a codebase - validated truths the AI should treat as baseline, optionally filtered by category.

• wyrm_truth_set

  Use to lock in a confirmed fact or constraint as project ground truth - architecture, conventions, key decisions. Injected first into every brief; re-setting the same category+key supersedes (history kept).

• wyrm_usage

  View token usage stats, cache hit rates, and estimated cost savings. Helps monitor and optimize AI credit consumption.

• wyrm_vector_setup

  Configure embedding provider and download models. Supports local, OpenAI, and Ollama.

• Code scan218 source files scanned15/25
• –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 6 days ago15/15
• Maintainer identitynamespace and repository owner differ; website matches verified namespace6/10

Grade history

• 2026-09-19restoreC → Bscore 74: npm install lifecycle script present; npm install lifecycle script present

Wyrm: common questions

Is Wyrm MCP server safe?

Mostly: it is graded B (74/100). Read the Wyrm safety report

How do I install Wyrm?

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

Does Wyrm need an API key?

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

Is Wyrm maintained?

The last commit was 6 days ago (2026-09-14). The latest release is v9.2.3.

What can I use instead of Wyrm?

Servers from other publishers that do the same job: Plur MCP server, Vestige MCP server and Prism Coder MCP server. Compare all Wyrm alternatives.