Community Edition
One binary.
Full retrieval.
Source-available retrieval engine with cryptographic receipts baked in. Append-only event store, BM25 search, 27 MCP tools, CROWN receipts, and optional dense vector retrieval. Run locally, in Docker, or on bare metal.
Up and running in 60 seconds
Start the server
One command. Runs on port 14800 (HTTP) and 14801 (MCP).
docker compose up -dStore a fact
Structured knowledge with confidence scores and versioning.
curl -X PUT http://localhost:14800/v1/facts \
-H "Content-Type: application/json" \
-d '{"entity":"project","key":"status","value":"Phase 1 complete"}'Connect your agent
Built-in MCP server with 27 tools. Any MCP-compatible client.
curl -X POST http://localhost:14801/mcp \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'What's inside
Click any card to learn more
Agent Passport
PrimaryIdentity, lineage, and reputation
Every agent earns a passport with a 5-tier reputation system. Sync operations require a minimum tier, so agents must prove local track record before sharing knowledge.
Tiers: unverified, basic (10+ receipts), established (100+), trusted (500+), elite (2000+). Pull requires basic tier. Push requires established. Passports track sponsor lineage and are BLAKE3-hashed for integrity. Tier upgrades happen automatically as receipt count grows.
The passport is the anchor of six identity layers
An agent answers six questions about itself. The passport provides identity; the other five layers plug in alongside it.
| Layer | Question | Source |
|---|---|---|
| Identity | Who am I to this platform? | Passport (exists) |
| Operator | Who do I work for and how do they work? | OperatorProfile (v2.3, shipped) |
| Directive | What am I here to do, specifically? | Task context / role / resume intent |
| Rules | What must I not do? | Constraints + Shield grants (exist) |
| Playbook | How is this work typically done here? | Agent Skills + Operating Rhythms |
| Continuity | What happened since last time? | resume_session + domain changelog |
MCP Server
27 built-in tools
Retrieval, facts, sessions, constraints, decisions, handoffs, and sync, all via the Model Context Protocol. Connect any agent.
CROWN Receipts
Cryptographic verification
Ed25519-signed verification receipts on every operation. Tamper-evident, independently verifiable, with full audit trails.
Fact Store
Structured knowledge with versioning
Entity/key/value store with confidence scores, automatic versioning, soft-delete, and agent-scoped privacy.
BM25 Search
Full-text retrieval
Text search with companion .ccxi indexes. Token budgets let you cap result size for 60–80% cost reduction vs top-K.
Bring Your Own Embeddings
You choose the model and hardware
Connect Ollama, vLLM, TEI, llama.cpp, or any OpenAI-compatible endpoint. Dense vector retrieval with the model you trust.
Constraint Governance
Policy enforcement for agents
Declare organisational constraints (boundaries, policies, relationships) and check proposed actions against them before executing.
What's included
Community Edition vs Optional vs Proprietary
| Feature | CE | Optional | Proprietary |
|---|---|---|---|
| Append-only event store (BLAKE3) | yes | ||
| CROWN receipts (Ed25519) | yes | ||
| Fact store + session store | yes | ||
| Built-in MCP server (27 tools) | yes | ||
| Prometheus /metrics | yes | ||
| HTTP + gRPC APIs | yes | ||
| CLI tooling (verify, replay, etc.) | yes | ||
| BM25 text search | yes | ||
| Agent Passport (5-tier reputation) | yes | ||
| Sync gated by passport tier | yes | ||
| Dense vector retrieval | BYOM | ||
| Graph signal fusion | hosted | ||
| GPU/CUDA acceleration | hosted | ||
| Remote sync (hosted platform) | hosted |
All 27 MCP tools
Grouped by category. Every tool your agents get out of the box.
Retrieval
3querySearch the retrieval index using BM25 + graph fusion. Returns scored results with query coverage. Use token_budget to cap results by token count.
query_scanLightweight scan returning metadata only (no content). Useful for checking what exists before expanding. Returns scores and token counts per result.
query_expandExpand previously retrieved results by segment or doc IDs to get full content.
Fact Store
6store_factStore a key-value fact against an entity. Facts carry optional receipt references and confidence scores. Set private to scope the fact to your agent identity only.
query_factsQuery the fact store by keyword, entity, or both. Results are ranked by confidence. Private facts are visible only to their owning agent.
delete_factSoft-delete a fact by its ID. The fact's receipt is preserved for audit.
list_entitiesDiscover all entity names in the fact store. Returns a sorted, deduplicated list of entity names with at least one active fact.
get_bootstrapQuery bootstrap knowledge at runtime. Returns facts stored under the __bootstrap__ entity prefix for onboarding, patterns, and error resolution.
fact_historyReturn the full version chain for a given (entity, key) pair. Shows how a fact evolved over time, including superseded versions.
Sessions
4get_sessionRetrieve your session state by ID. Authenticated agents see only their own session namespace.
save_sessionCreate or update your session state. Supports optional time-to-live for automatic expiry.
list_sessionsList active session IDs visible to you. Returns a sorted list.
delete_sessionDelete one of your sessions by ID.
Constraints
3declare_constraintDeclare an organisational constraint (boundary, relationship, policy, or context flag) that agents must respect. Stored with severity levels.
get_constraintsList active organisational constraints, optionally filtered by type or status. Sorted by severity.
check_constraintsCheck a proposed action against all active constraints. Returns a verdict: pass, warn, or block based on keyword matching against constraint assertions.
Decisions
1record_decisionRecord why a decision was made. Stores an append-only, BLAKE3-hashed decision record as a fact. Queryable via query_facts.
Handoff
2create_handoffPackage session state and relevant non-private facts into a server-authenticated handoff bundle for another agent.
accept_handoffAccept a server-authenticated handoff package from another agent, restoring session state and facts into your namespace.
Passport
2issue_passportIssue an agent passport for the calling agent. Tracks identity, sponsor lineage, and reputation tier. Required before sync operations.
get_passportReturn the calling agent's passport with current reputation tier and receipt count. Automatically upgrades tier when new thresholds are met.
Sync
3sync_pullPull latest facts from a remote CoreCrux instance. Requires basic passport tier (10+ receipts). Resumes from the last pull cursor.
sync_pushPush local facts to a remote CoreCrux instance. Requires established passport tier (100+ receipts). Private facts are never pushed.
sync_statusShow sync mode: local-only, manual sync, full background sync, or degraded. No passport required.
Observability
1get_gapsRetrieve known knowledge gaps from the ops observation layer. Check after low-coverage queries to understand what is missing.
Agent
1get_agent_identityReturn the calling agent's name. Returns "anonymous" if no agent identity is authenticated.
Update
1update_statusShow whether this checkout is current, behind, ahead, diverged, disabled, or unavailable relative to the tracked git branch.
Ready to run your own retrieval engine?
Source-available. No credit card. No cloud dependency.