Skip to content

Connect your coding agent ​

Each coding agent learns about you from scratch. Connect them to Configure once and, with your approval, they share one context. It holds your stack, your conventions, and the lessons one agent learned last week.

Setup has two halves. People often skip the second one.

Memory is the MCP server. It gives the agent the configure_* tools.

Habit is the instruction file that tells the agent to use them. Without it, the tools sit unused. A model calls a tool when it wants something, and it never decides on its own to load its context. Product agents show the same measured behavior.

Install ​

Run both commands. The habit instructions describe your memory. Append them to a global rules file. Do not commit them to a shared team repo, where they would apply to people who never opted in.

The Cursor and Windsurf commands below append to AGENTS.md in the current folder. Run them only in a private checkout. For a shared repo, paste agents-block.md into your personal rules instead.

bash
# 1. Memory
claude mcp add --transport http configure https://mcp.configure.dev

# 2. Habit (installs the configure-memory skill and a session-start digest hook)
curl -fsSL https://docs.configure.dev/skill/install.mjs | node -
bash
# 1. Memory
codex mcp add configure --url https://mcp.configure.dev
codex mcp login configure

# 2. Habit (idempotent: skips if already present)
mkdir -p ~/.codex && grep -qs "User memory (Configure)" ~/.codex/AGENTS.md \
  || curl -fsSL https://docs.configure.dev/skill/agents-block.md >> ~/.codex/AGENTS.md
bash
# 1. Memory (add -s user for a global server)
gemini mcp add --transport http configure https://mcp.configure.dev

# 2. Habit (Gemini CLI reads GEMINI.md by default)
mkdir -p ~/.gemini && grep -qs "User memory (Configure)" ~/.gemini/GEMINI.md \
  || curl -fsSL https://docs.configure.dev/skill/agents-block.md >> ~/.gemini/GEMINI.md
bash
# 1. Memory: add to ~/.cursor/mcp.json (Settings -> MCP), then approve the OAuth popup
#    { "mcpServers": { "configure": { "url": "https://mcp.configure.dev" } } }

# 2. Habit (Cursor reads AGENTS.md natively)
grep -qs "User memory (Configure)" AGENTS.md \
  || curl -fsSL https://docs.configure.dev/skill/agents-block.md >> AGENTS.md
bash
# 1. Memory: add to ~/.codeium/windsurf/mcp_config.json (Settings -> Cascade -> MCP)
#    { "mcpServers": { "configure": { "serverUrl": "https://mcp.configure.dev" } } }

# 2. Habit
grep -qs "User memory (Configure)" AGENTS.md \
  || curl -fsSL https://docs.configure.dev/skill/agents-block.md >> AGENTS.md

Approve the OAuth prompt the first time. The configure_* tools then appear in the agent's tool list. Repeat this for every agent you use. The same context follows you into each one.

For Codex there is a second way in: npx -y @configure-ai/mcp login --personal --client codex approves in your browser and installs a local Configure server instead of Codex's own OAuth. See Connect over MCP.

What the Claude Code installer changes ​

The installer writes the configure-memory skill to ~/.claude/skills/configure-memory and changes ~/.claude/settings.json in two ways:

  • It registers a session-start hook. The hook reads the Configure token Claude Code already stored when you approved the OAuth prompt, and loads a short digest of your context. It reads that token from the macOS keychain item Claude Code-credentials, or else from ~/.claude/.credentials.json. Set CONFIGURE_TOKEN, and optionally CONFIGURE_MCP_URL, to use your own token instead of both. It never mints a credential of its own.
  • It adds four tools to permissions.allow, so they run without asking: configure_profile_read, configure_profile_search, configure_profile_remember, and configure_profile_commit. configure_profile_forget and configure_profile_import still ask. The installer prints the list when it finishes.

To remove the skill, the hook, and the permissions the installer added (a tool you had allowed yourself stays allowed, except after an install made before November 2026, which did not record what it added, so all four are removed):

bash
curl -fsSL https://docs.configure.dev/skill/install.mjs | node - --uninstall

It leaves the MCP server in place. Remove that with claude mcp remove configure.

Verify it worked ​

Ask the agent something only your context knows:

what test runner do I normally use?

A connected agent calls configure_profile_search and answers from the result. If the agent says it has no way to know, it has memory but is missing the habit. Re-run the second command.

Hand work between agents ​

Once more than one agent is connected, they can pass work to each other through a shared project shelf. You do not have to paste transcripts. See Projects.

Infrastructure to connect agents, apps, and chats