Persistent project context for xAI Grok. IANA-registered .faf format.
Scored across 20 tools
A large cluster of tools overlaps heavily around scoring/validating .faf content: faf_score, faf_validate, faf_analyze (a 'one-call composite' returning score+valid+tier), faf_get_tier, faf_gate, refresh_faf, and even faf_orchestrate_recommendation all touch the same scoring substrate. faf_analyze effectively subsumes faf_score/faf_validate/faf_get_tier, creating real misselection risk. The retrieval tools (faf_section, search_context, faf_collections_search, faf_memory) are better delineated, but the scoring cluster drags the score down.
All names are snake_case, which is consistent, but the convention is split between a 'faf_' namespace (faf_score, faf_gate, faf_analyze) and unprefixed verb_noun tools (get_soul, list_souls, search_by_tag, generate_faf_from_github) plus vague noun-style names (delta_check, tag_intel). The mix of prefixed and unprefixed families within one server is readable but not uniform.
20 tools sits in the heavy band (16-25) for what is essentially a context/scoring server. Several scoring variants and search variants exist that could be consolidated, so not every tool clearly earns its place. It is still navigable, but borders on over-scoped.
The surface covers scoring, validation, gating, generation (from GitHub), soul retrieval, tag/search, and token estimation well. However, souls appear read-only — there is no create_soul, update_soul, or delete_soul, so an agent cannot persist or edit context souls despite the 'persistent project context' purpose. Notable write-side gaps remain.
Determine if a topic needs FULL, DELTA, or X-DELTA soul.
Inputs: topic (required).
One-call composite — returns score, tier-ready, valid, and engine identifier. Two WASM calls, sub-millisecond total.
Inputs: content (required).
Print a business card for an agent from its own name, description, colour and picture. Returns the agent's .fafa (FAF Agent file) and a link that opens its card. Nothing is stored: the .fafa travels in the link. The card says "self-declared" because nothing is checked against a repo.
Inputs: name (required), icon, accent, description.
Phase III (FRC §7b) — semantic search over a Grok Collection at the edge, KV-cached (1h TTL). Returns matched chunks (content, score, file). Requires the XAI_API_KEY secret; composes with faf_section (structural) for hybrid retrieval. Handled env-aware in the MCP handler (needs the key + KV).
Inputs: query (required), collection_id (required), limit.
Estimate token count for arbitrary content via the Zig WASM engine. Sub-millisecond, zero allocations. Useful for context-budget planning.
Inputs: content (required).
Phase III (FRC) — pre-promotion quality gate. Scores .faf content (edge Mk4) + estimates tokens and returns a deterministic promote/hold verdict BEFORE it goes to a Grok Collection. Promote IFF score >= min_score AND tokens <= max_tokens (defaults 85/8000). Edge parity with the local gate; the hold-hint can't list empty slots at the edge.
Inputs: content (required), min_score, max_tokens.
Resolve the FAF tier for a given numeric score. Returns the tier symbol (Trophy/Gold/Silver/Bronze/etc.) per the canonical tier-table.
Inputs: score (required).
Phase III (FRC) — query the durable .fafm model by type/tag/priority/text. Omit filters for a structured summary. .fafm is NOT scored: this SELECTS facts (provenance preserved), never grades them.
Inputs: content (required), tag, type, query, +1 more.
Takes raw content strings (.faf, .fafm, and optionally package.json/CHANGELOG.md/README.md) and runs deterministic drift + contradiction signals across the FAF substrate. Returns a structured Recommendation (recommend, severity, reason, summary) with hints containing the current effective_policy and partial[] for any stateful signals unavailable on the current surface. Light-lane execution (hosted) is WASM-pure with no filesystem access. Heavy-lane execution (local via bunx/rust-faf-mcp) has full FS + persisted state. Advisory only — never auto-fires.
Inputs: faf, fafm, readme, changelog, +1 more.
Score .faf YAML content with the always-33 faf-kernel (WASM). Returns 0-100 (capped). The same score faf-cli and every FAF MCP give. Runs at the edge.
Inputs: content (required).
Phase III (FRC) — returns an EXACT, WHOLE .faf section by dotted path (e.g. "stack", "human_context"), structure preserved — the deterministic complement to blind chunking. Omit "section" to list every path.
Inputs: content (required), section.
Validate .faf YAML content: true if mission-ready (always-33 score of 100).
Inputs: content (required).
Generate a .faf file from any public GitHub repository WITHOUT cloning. Extracts 6 Ws from README, analyzes stack from languages and package.json, and generates Championship-grade AI context. Returns .faf content, quality score, and metadata.
Inputs: repo (required).
Fetch a context soul by name. Returns structured AI context.
Inputs: soul (required).
List context souls: showcase souls first, then claimed souls. Paged: pass the returned cursor for more.
Inputs: limit, cursor.
List all unique tags used in a soul, with counts.
Inputs: soul (required).
Re-ground on .faf content — re-score with the always-33 faf-kernel (33 slots; slotignored slots drop out), report drift vs an optional baseline score, and return a stamped re-ground. The explicit re-grounding primitive for long sessions: drift → refresh → re-grounded. Built for Grok, by request.
Inputs: content (required), baseline, verbatim.
Find all entries in a soul with a specific tag.
Inputs: tag (required), soul (required).
Full-text search across souls (showcase first, then claimed souls; [private] sections are never searched). Returns matching lines only. Searches a page of souls per call: pass the returned cursor for more.
Inputs: query (required), soul, limit, cursor.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Persistent project context for Claude. IANA-registered .faf format.
Persistent project context — Rust-native MCP server. IANA-registered .faf format.
Your portable context layer — load it into any AI.
Project memory, documentation and controlled secret access for AI coding agents.
Add one secure layer between your agents and this server.
