MCP Server

Your Brain doesn't have to stay inside Agently. The Agently Brain MCP server exposes your company's temporal knowledge graph to any client that speaks the Model Context Protocol — so Claude Code, Claude Desktop, claude.ai, Cursor, VS Code, ChatGPT, and any other MCP-capable agent can use your Brain as their brain.

https://api.agently.dev/mcp

What your agent gets

Tool	What it does
`search`	Hybrid semantic + keyword search — ranked facts with temporal validity and entity references
`find_entity`	Resolve a name ("Acme", "Jane") to graph entities with UUIDs and summaries
`get_neighbors`	Walk one hop from an entity — every fact touching it, plus connected entities
`get_schema`	Brain overview: counts, most-connected entities, source kinds, data recency
`recent_episodes`	Newest ingested content, with previews
`remember`	Store a durable fact in the Brain (write-scoped credentials only)

Two things make this different from pointing your agent at a folder of documents:

It's a graph. Search results carry the entities on every fact, so an agent can pivot from "found a fact about Acme" to "show me everything connected to Acme" without re-searching.
It's temporal. Facts have validity windows. A fact marked SUPERSEDED was true once and later contradicted — agents can tell current truth from history.

Full parameter details live in the MCP Tools Reference.

Quickstart

1. Get a credential

In Agently, go to Settings → API keys → New key. Name it after the client ("Claude Code — laptop"), optionally enable Allow writes if the agent should be able to store facts, and copy the agently_sk_... key — it's shown once.

Clients that support MCP OAuth (Claude Desktop connectors, claude.ai, ChatGPT) don't need a key at all — they discover Agently's OAuth flow automatically and walk you through a consent screen instead. See MCP Authentication & API.

2. Connect your client

Claude Code

claude mcp add --transport http agently-brain https://api.agently.dev/mcp \
  --header "Authorization: Bearer agently_sk_..."

Or commit it for your whole team in the project's .mcp.json (use an env var for the key):

{
  "mcpServers": {
    "agently-brain": {
      "type": "http",
      "url": "https://api.agently.dev/mcp",
      "headers": { "Authorization": "Bearer ${AGENTLY_API_KEY}" }
    }
  }
}

Cursor — Settings → MCP → Add new MCP server, or in .cursor/mcp.json:

{
  "mcpServers": {
    "agently-brain": {
      "url": "https://api.agently.dev/mcp",
      "headers": { "Authorization": "Bearer agently_sk_..." }
    }
  }
}

VS Code / GitHub Copilot — in .vscode/mcp.json:

{
  "servers": {
    "agently-brain": {
      "type": "http",
      "url": "https://api.agently.dev/mcp",
      "headers": { "Authorization": "Bearer agently_sk_..." }
    }
  }
}

Claude Desktop / claude.ai — Settings → Connectors → Add custom connector → enter https://api.agently.dev/mcp. The OAuth consent flow opens in your browser: sign in, pick a workspace, approve. No key needed.

ChatGPT — Settings → Apps & Connectors → Add → enter https://api.agently.dev/mcp and complete the OAuth flow.

3. Ask something

"Search the brain for our Q3 pricing decisions."
"What does the brain know? Give me an overview."
"Find the entity 'Acme Corp' and show me everything connected to it."
"Remember that we chose Postgres over MySQL for the analytics service." (write scope)

A good agent chains the tools on its own: search → spot an interesting entity → get_neighbors to expand it. The server's tool descriptions steer that behavior.

How it stays safe

Every credential is pinned to one workspace, chosen when it's created. No tool takes a workspace parameter — an agent can never reach across workspaces.
Credentials are read-only by default. The remember write tool only exists for keys or tokens granted write scope; read-only clients never even see it.
The server is stateless (MCP spec 2026-07-28): no sessions, nothing persisted between requests beyond the Brain itself.

Agent-readable versions of these docs: llms.txt · llms-full.txt