hebbrix-mcp 0.3.1

Hebbrix MCP server — long-term memory and knowledge graph for any MCP-compatible agent (Claude Desktop, Cline, Cursor, etc.)

hebbrix-mcp

Long-term memory and a knowledge graph for any MCP-compatible agent.

Your agent forgets everything when the session ends. This server fixes that — and goes further than a plain memory store:

• Memory — store, search, correct, and version facts across sessions
• Knowledge graph — entities, relationships, timelines, and "what was true at time X"
• Reasoning — ask how confident the agent should be before acting, and log outcomes so it improves

Works with Claude Desktop, Claude Code, Cursor, Cline, Continue, and any other MCP client. Backed by Hebbrix.

---

Quick start (10 seconds, no account)

1. Install

pip install hebbrix-mcp

2. Add to your MCP client — no API key needed:

Claude Desktop — ~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "hebbrix": { "command": "hebbrix-mcp" }
  }
}

Claude Code

claude mcp add hebbrix -- hebbrix-mcp

Cursor — ~/.cursor/mcp.json

{
  "mcpServers": {
    "hebbrix": { "command": "hebbrix-mcp" }
  }
}

Cline / Continue / other

Point your MCP servers config at the hebbrix-mcp command (stdio). Same shape as above.

3. Restart the client. Done — your agent now has persistent memory.

What just happened?

On first run with no API key, the server mints a free agent account automatically (no email, no dashboard, ~1 second) and saves the credentials to ~/.hebbrix/config.json. The account includes:

Learning events (writes)	300
Retrievals (searches)	2,000
Expiry	14 days after last activity, if unclaimed

Every tool result includes a hebbrix_usage block (tier, usage, expiry), so the agent always knows where it stands and will tell you when it's time to claim.

Keep it forever (one command)

hebbrix-mcp claim --email [e-mail]

You'll get a 6-digit code by email; enter it and the account switches to the free monthly tier with no expiry. Same API key, all memories carry over. Confirming decision outcomes (hebbrix_log_decision) also extends the trial — the system rewards exactly the usage that makes it smarter.

---

Using your own API key

Already have a Hebbrix account? Get a key at hebbrix.com/dashboard/api-keys and pass it instead:

{
  "mcpServers": {
    "hebbrix": {
      "command": "hebbrix-mcp",
      "env": {
        "HEBBRIX_API_KEY": "mem_sk_...",
        "HEBBRIX_COLLECTION_ID": "your-default-collection-uuid"
      }
    }
  }
}

The env var always wins over saved agent-mode credentials.

---

Tools

15 tools, one resource, one prompt. A server-level instruction block teaches the model when to reach for each, so a well-behaved agent searches before answering and remembers what matters — without being told.

Memory

Tool	What it does
hebbrix_remember(content, tags?, collection_id?, verbatim?)	Store a fact, decision, or preference. verbatim=true skips fact-extraction
hebbrix_search(query, limit?, collection_id?)	Semantic search (hybrid vector + BM25 + graph retrieval)
hebbrix_get(memory_id)	Fetch one memory with metadata
hebbrix_update(memory_id, content?, importance?)	Correct a memory in place — old versions are kept
hebbrix_forget(memory_id)	Delete a memory
hebbrix_list(limit?, collection_id?)	List recent memories
hebbrix_history(memory_id)	See how a memory changed over time

Knowledge graph

Reads are available on every tier (including agent mode). Graph writes and inference need a Pro plan.

Tool	What it does
hebbrix_search_entities(entity_type?, limit?, collection_id?)	List known entities (people, orgs, tools, places)
hebbrix_entity_timeline(entity_name, collection_id?)	What was true about an entity, and when
hebbrix_graph_query(query?, entity?, relation_type?, depth?, timestamp?)	Query relationships — pass a timestamp to ask about a point in time
hebbrix_contradictions(memory_id?)	Surface facts that conflict with each other

Reasoning & account

Tool	What it does
hebbrix_confidence(query, collection_id?)	How confident should the agent be before acting? Grounded in memory + past outcomes
hebbrix_log_decision(description, outcome?, decision_type?)	Record a decision and how it turned out — feeds future confidence
hebbrix_list_collections()	List the memory spaces this key can use
hebbrix_account_status()	Tier, usage, limits, and expiry

Also: the hebbrix://profile resource and the context prompt inject the user's compiled profile into the conversation.

---

Running modes

1. Local (default) — stdio

What the quick start above does. One process per client, credentials from env or ~/.hebbrix/config.json.

2. Self-hosted HTTP — one instance, your machines

HEBBRIX_API_KEY=mem_sk_... hebbrix-mcp --transport streamable-http
# serves http://127.0.0.1:8080/mcp

{ "mcpServers": { "hebbrix": { "url": "http://127.0.0.1:8080/mcp" } } }

3. Hosted multi-tenant — one instance, many users

The server holds no key at all; every request authenticates with its own Authorization header:

HEBBRIX_MCP_MULTI_TENANT=1 HEBBRIX_MCP_HOST=0.0.0.0 hebbrix-mcp --transport streamable-http

{ "mcpServers": { "hebbrix": {
  "url": "https://your-host/mcp",
  "headers": { "Authorization": "Bearer mem_sk_..." }
}}}

In this mode there is no default collection — pass collection_id on tool calls.

---

Configuration

All optional. With nothing set, the server starts in agent mode.

Variable	Default	Purpose
HEBBRIX_API_KEY	(agent mode mints one)	Your Hebbrix bearer token
HEBBRIX_COLLECTION_ID	(agent mode sets one)	Default collection for writes/reads
HEBBRIX_API_BASE	https://api.hebbrix.com/v1	API endpoint override
HEBBRIX_CONFIG	~/.hebbrix/config.json	Where agent-mode credentials are saved
HEBBRIX_MCP_HOST	127.0.0.1	Bind host (HTTP transports)
HEBBRIX_MCP_PORT	8080	Bind port (HTTP transports)
HEBBRIX_MCP_MULTI_TENANT	off	Hosted mode: per-request header auth

---

How it works

┌──────────────────┐   MCP (stdio or HTTP)   ┌─────────────┐    HTTPS     ┌──────────┐
│ Claude / Cursor / │ ───────────────────────→│ hebbrix-mcp │─────────────→│ Hebbrix  │
│ Cline / any agent │      tool calls         │   (this)    │   REST API   │  cloud   │
└──────────────────┘                          └─────────────┘              └──────────┘

This package owns zero state. Tool calls become REST calls against your Hebbrix account; memories, embeddings, the knowledge graph, and retrieval all live in the Hebbrix backend. Delete this package and your memories are still there.

Limits degrade gracefully

Agent-mode accounts never break mid-task. When a limit is reached you get a structured error, not a failure:

Code	Meaning	What still works
WRITE_LIMIT_REACHED	300 lifetime writes used	Reads, searches, confirmations
READ_LIMIT_REACHED	2,000 lifetime retrievals used	Claim to continue
SHADOW_READ_ONLY	Unclaimed past 14 days	Reads (7-day grace window)
SHADOW_EXPIRED	Past the grace window	Nothing — account is reaped
CLAIM_REQUIRED_FOR_BATCH	Batch writes need a claimed account	Everything else

Every error carries a resolve field with the exact command to fix it, so agents can relay it to you verbatim.

---

Troubleshooting

"HTTP 401" on every call — the key is wrong or revoked. Unset HEBBRIX_API_KEY, delete ~/.hebbrix/config.json, and restart to re-provision; or paste a fresh key from the dashboard.

Agent mode won't start (auto-signup unavailable) — signup may be at daily capacity (it's capped) or your network blocks the API. Set HEBBRIX_API_KEY from the dashboard instead.

claim says EMAIL_IN_USE — v1 claiming needs an email with no existing Hebbrix account. Use a fresh address (a [e-mail] alias works).

A memory doesn't show up in search immediately — indexing is asynchronous; typical convergence is under 30 seconds.

Multi-tenant mode returns errors about collections — there's no default collection in hosted mode; pass collection_id explicitly.

---

Development

git clone https://github.com/Hebbrix/hebbrix-mcp
cd hebbrix-mcp
./quick_setup.sh            # venv + editable install
source venv/bin/activate
pytest tests/ -q            # 11 offline tests, no network needed
hebbrix-mcp                 # starts in agent mode on stdio

See CHANGELOG.md for release history and CONTRIBUTING.md for how to contribute.

License

MIT — see LICENSE.

Related

• Hebbrix documentation
• MCP integration guide
• Model Context Protocol