MCP Setup
Loomtracer’s command-line tool, loomtracer, exposes its full decision-graph surface as an MCP server. Connect your agents
to your self-hosted cell over HTTP (a shared, team-wide decision graph), or run the
local stdio server yourself for single-user use.
Self-hosted cell — HTTP
Section titled “Self-hosted cell — HTTP”Your self-hosted cell serves MCP at /mcp. Agents
connect to that endpoint and write to a shared, team-wide decision graph — no local
binary required on the agent’s machine. Authentication is a bearer token:
provision a tenant and use the
lt_tk_... token it returns. The examples below use http://localhost:8080/mcp;
substitute your cell’s address.
Claude Code
Section titled “Claude Code”Add to .mcp.json in your project root:
{ "mcpServers": { "loomtracer": { "type": "http", "url": "http://localhost:8080/mcp", "headers": { "Authorization": "Bearer lt_tk_..." } } }}Or from the CLI:
claude mcp add --transport http loomtracer http://localhost:8080/mcp \ --header "Authorization: Bearer lt_tk_..."Reload Claude Code and all loomtracer tools are available.
Claude Desktop
Section titled “Claude Desktop”Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or
%APPDATA%\Claude\claude_desktop_config.json (Windows):
{ "mcpServers": { "loomtracer": { "type": "http", "url": "http://localhost:8080/mcp", "headers": { "Authorization": "Bearer lt_tk_..." } } }}Cursor
Section titled “Cursor”Add to ~/.cursor/mcp.json or the project-level .cursor/mcp.json:
{ "mcpServers": { "loomtracer": { "url": "http://localhost:8080/mcp", "headers": { "Authorization": "Bearer lt_tk_..." } } }}Put the token in an environment variable and register the cell:
export LOOMTRACER_TOKEN=lt_tk_...codex mcp add loomtracer --url http://localhost:8080/mcp --bearer-token-env-var LOOMTRACER_TOKENCodex writes the server to ~/.codex/config.toml and reads the bearer token from that variable.
Local stdio MCP
Section titled “Local stdio MCP”For single-user use, run the local stdio server. The server is a thin transport over the same commands layer the CLI uses.
Start the server
Section titled “Start the server”loomtracer --loomtracer-dir ./loomtracer/ mcpThe server reads from stdin and writes to stdout in the MCP protocol format.
Claude Desktop (local)
Section titled “Claude Desktop (local)”{ "mcpServers": { "loomtracer": { "command": "loomtracer", "args": ["mcp"], "env": { "LOOMTRACER_DIR": "/absolute/path/to/your/loomtracer/dir" } } }}Claude Code (local)
Section titled “Claude Code (local)”Add to .mcp.json in the project root:
{ "mcpServers": { "loomtracer": { "command": "loomtracer", "args": ["--loomtracer-dir", "./loomtracer/", "mcp", "--agent-tool", "claude"] } }}Or set LOOMTRACER_DIR and omit --loomtracer-dir:
{ "mcpServers": { "loomtracer": { "command": "loomtracer", "args": ["mcp", "--agent-tool", "claude"], "env": { "LOOMTRACER_DIR": "./loomtracer/" } } }}--agent-tool claude files captures under Claude Code (see Who a capture is filed under).
Cursor (local)
Section titled “Cursor (local)”{ "mcpServers": { "loomtracer": { "command": "loomtracer", "args": ["mcp", "--agent-tool", "cursor"], "env": { "LOOMTRACER_DIR": "/path/to/loomtracer/" } } }}Codex (local)
Section titled “Codex (local)”codex mcp add loomtracer -- loomtracer mcp --agent-tool codexThis writes the server to ~/.codex/config.toml; codex mcp list shows it enabled.
[mcp_servers.loomtracer]command = "loomtracer"args = ["mcp", "--agent-tool", "codex"]Available tools
Section titled “Available tools”The loomtracer MCP server exposes 40 tools. See MCP Tools reference
for full parameter documentation.
| Tool | Type | Description |
|---|---|---|
capture_decision | write | Record a decision with rationale and options |
capture_evidence | write | Record supporting evidence for a decision |
capture_hypothesis | write | Record a hypothesis still in flight |
disagree_decision | write | Contest a decision as an actor |
supersede_decision | write | Supersede a prior decision with a new one |
move_decision | write | Move a decision to another project, recorded with who, when, from, to and why |
retitle_decision | write | Rename a decision; the old title stays readable |
link_supersession | write | Record that one already-recorded decision replaces another, keeping its own decider |
ground_decision | write | Say what an existing decision rests on, after the fact |
reaffirm_decision | write | Say a decision still holds after something it rests on changed |
request_decision | write | Ask a question before any decision answers it |
acknowledge_suggestion | write | Record that a finding from get_suggestions was looked at |
get_decision | read | Retrieve a decision by ID with derived status |
get_relevant_decisions | read | Search by topic, status, actor, or time window |
get_situational_decisions | read | Decisions bearing on a situation — matches by topic or evidence overlap, no ID needed |
get_supersession_chain | read | Walk the full supersession history backward |
get_decision_neighborhood | read | One-hop graph neighborhood around a decision — actors, options, evidence, hypotheses |
search_decisions | read | Full-text search across the ledger |
recent_decisions | read | List recently proposed decisions |
dump_graph | read | Export the full projected graph (DOT or JSON) |
get_decision_outcome | read | Whether a decision still holds — superseded, stale premises, or contested |
get_reexamine_items | read | Decisions resting on something that changed that nobody has looked at again |
get_waiting_requests | read | Questions asked that no decision answers yet, oldest first |
get_contested_decisions | read | Decisions in contest: who accepted and who rejected, or conflicting answers |
get_changed_decisions | read | Decisions revised or superseded in a window, with what changed and who changed it |
get_open_why_decisions | read | Binding decisions whose why is still to answer |
loomtracer_compact_view | layer-3 | Compact summary of a decision and its context |
recall_decisions | layer-3 | Search + ranked digest — answers “what was decided about X?” |
summarize_decisions | layer-3 | Concise text summary of one or more decisions |
decision_quality_candidates | layer-3 | Bulk quality-signal pull for external scorers |
get_decision_context | layer-3 | Context record for a decision — authorship, review depth, evidence richness |
decision_context_candidates | layer-3 | Bulk context-feature pull, the independent-variable side of quality scoring |
score_decision | layer-3 | The quality profile of one decision: seven dimensions, no score, no tier |
scan_decision_quality | layer-3 | Attention findings: decisions that need a look, such as a bet past its check date |
get_suggestions | layer-3 | The attention findings nobody has acknowledged yet |
rank_decisions_by_importance | layer-3 | Which decisions carry the most impact, each with who decided it |
scan_misfiled_decisions | layer-3 | Flag decisions carrying a caller-named “foreign” topic key — report only, never moves anything |
analyze_failure_modes | layer-3 | Aggregate failure-rate patterns across authorship, review, and context |
classify_queue_list | read | List pending ingest batches with turn text, plus today’s classification budget (HTTP transport only) |
classify_queue_submit | write | Submit captures for one or more pending batches as one classification event (HTTP transport only) |
Who a capture is filed under (local stdio)
Section titled “Who a capture is filed under (local stdio)”A capture call may name its actor in actor_id; it is optional. Without one, the server records
agent:<tool>:mcp-<session-id>, and a capture that names no project lands in that agent’s
personal project. <tool> comes from --agent-tool on the server’s command line, so name your
client there:
loomtracer mcp --agent-tool claude # Claude Codeloomtracer mcp --agent-tool codex # Codexloomtracer mcp --agent-tool cursor # CursorWithout the flag the server usually cannot tell which client started it, and files captures under
codex. The server records source=agent and a per-session source_ref for every write.
Next steps
Section titled “Next steps”- MCP Tools reference — full parameter documentation for all 40 tools
- Agent Capture guide — how agents capture decisions automatically
- Quickstart — install the binary, or run your own cell with Docker