MCP Portal
Sign in ← LTHN

OpenBrain

openbrain

Available

Shared agent knowledge store with vector search

About

Shared vector knowledge store for AI agents. Powered by Qdrant (semantic search),
Ollama (embeddings), and MariaDB (relational metadata). Agents store decisions,
observations, conventions, and research as memories that persist across sessions
and can be recalled using natural language queries.

All memories are workspace-scoped for multi-tenant isolation. Supports memory
supersession chains, confidence scoring, expiration, and project-level filtering.

Use when

  • • Storing decisions, observations, or conventions for other agents
  • • Searching knowledge with natural language queries
  • • Sharing context between agents across sessions
  • • Building a persistent knowledge base for a project
  • • Recalling what was decided or observed in previous sessions

Don't use when

  • • Storing ephemeral session data (use state_set instead)
  • • Simple key-value lookups (use state_get instead)
  • • Storing large binary files or attachments
  • • Real-time data that changes every request

Connection

Call tools on this server via HTTP: 

curl -X POST https://mcp.lthn.ai/tools/call \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "server": "openbrain",
    "tool": "brain_remember",
    "arguments": {}
  }'

Full setup guide →

Tools (4)

brain_remember

Store a memory in the shared OpenBrain knowledge store

Example prompts:

  • "Remember that we decided to use UUIDs for all primary keys"
  • "Store this architecture decision about the event bus"
  • "Save the convention that all API responses use snake_case"
Parameters
content * string

The knowledge to remember (max 50,000 characters)

type * string

Memory type classification

tags array

Optional tags for categorisation

project string

Optional project scope (e.g. repo name)

confidence number

Confidence level from 0.0 to 1.0 (default 0.8)

supersedes string

UUID of an older memory this one replaces

expires_in integer

Hours until this memory expires (null = never)

brain_recall

Semantic search across the shared OpenBrain knowledge store

Example prompts:

  • "What do we know about the database schema?"
  • "Recall any decisions about authentication"
  • "Find observations about performance issues"
Parameters
query * string

Natural language search query (max 2,000 characters)

top_k integer

Number of results to return (default 5, max 20)

filter object

Optional filters (project, type, agent_id, min_confidence)

brain_forget

Remove a memory from the shared OpenBrain knowledge store

Example prompts:

  • "Forget that memory about the old API format"
  • "Remove the incorrect architecture note"
Parameters
id * string

UUID of the memory to remove

reason string

Optional reason for forgetting this memory

brain_list

List memories without vector search

Example prompts:

  • "List all architecture memories"
  • "Show what agent-42 has stored"
  • "List recent decisions for the core-api project"
Parameters
project string

Filter by project scope

type string

Filter by memory type

agent_id string

Filter by originating agent

limit integer

Maximum results to return (default 20, max 100)

Workflows (2)

Knowledge Capture

Store and organise knowledge during a work session

  1. brain_remember — Store key decisions as they are made
  2. brain_remember — Record important observations
  3. brain_list — Review what has been stored

Context Recovery

Recall relevant knowledge when starting a new session

  1. brain_recall — Search for relevant prior knowledge
  2. brain_list — Review recent decisions
  3. Continue work — Use recalled context to inform current session
View Usage Analytics