MCP Server

Your Brain doesn't have to stay inside Agently. The Agently Brain MCP server gives any client that speaks the Model Context Protocol access to your company's saved knowledge and shared notes. Claude Code, Codex, Claude Desktop, claude.ai, Cursor, VS Code, ChatGPT and other MCP clients can use your Brain as their brain.

https://api.agently.dev/mcp

What your agent gets

Tool	What it does
`search`	Searches saved files, web pages, snippets and synced documents. Returns passages with their source and dates
`list_memories`	Lists the workspace's shared notes
`read_memory`	Reads one shared note
`remember`	Saves or updates a shared note (write access, owner or admin)
`forget`	Deletes a shared note or a remembered snippet (write access)
`prepare_brain_capture`	Shows an editable card to save a conversation's takeaways to the Brain. Nothing is saved until you click Save to Brain
`save_page`	Saves a document, HTML or React page, SVG, Mermaid diagram, code or CSV the client made as a private page (write access)
`list_skills`, `get_skill`	List the skills that are on in your workspace and fetch one as a portable SKILL.md

Search covers what is saved in the Brain. Live reads of your mail, Slack or CRM happen inside Agently, not over MCP. Private notes are never shared over MCP.

Reads are mostly free. The first 1,000 read calls per workspace each month cost nothing, so pointing your daily driver at the Brain doesn't burn your allowance.

Full parameters live in the MCP Tools Reference.

Quickstart

1. Get a credential

Create a workspace API key in the Agently app (see Agently API keys and MCP), or open Brain and choose Connect an agent. Name the key after the client ("Claude Code, laptop"). Turn on Allow writes if the agent should save notes, pages or reviewed captures. Copy the agently_sk_... key; it's shown once.

Clients that support MCP OAuth (Claude Desktop, claude.ai, ChatGPT) don't need a key. They walk you through a sign-in and 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}" }
    }
  }
}

Codex CLI

export AGENTLY_MCP_API_KEY='agently_sk_...'
codex mcp add agently-brain --url https://api.agently.dev/mcp --bearer-token-env-var AGENTLY_MCP_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, find Agently in Claude's connector directory and connect it. Sign in to Agently, pick a workspace and approve. No key needed.

ChatGPT, Settings → Apps & Connectors → Add → enter https://api.agently.dev/mcp and complete the sign-in.

3. Ask something

"Search the brain for our Q3 pricing decisions."
"What shared notes do we have about onboarding?"
"Remember that we chose Postgres over MySQL for the analytics service." (write access)
"Prepare this discussion for my team to review and save to Brain."
"Which of our Agently skills covers sales outreach?"
"Save this project's plan and key decisions to Agently." (write access; it becomes a private page in Pages)

Which way it works

The connection runs one way: your AI app uses your Brain. Jarvis can't open the app's own chats, projects or files. To have Jarvis work with something from ChatGPT or Claude, ask that app to save it to Agently, then ask Jarvis about the page it saved.

The server also offers prompts that show up as slash commands in some clients: catch_me_up, company_context, recall and capture_for_team, plus one per skill active in your workspace.

How it stays safe

Every credential is pinned to one workspace, chosen when it's created. No tool takes a workspace parameter, so an agent can never reach across workspaces.
Credentials are read-only by default. Write tools only appear for keys or tokens with write access.
Changing shared notes also needs the credential's owner to be a workspace owner or admin. If they leave the workspace, the credential stops working.
The server is stateless: no sessions, nothing kept between requests beyond the Brain itself.
forget only removes shared notes and remembered snippets. Uploaded files and content synced from your tools are out of its reach.

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