# Coding agents

Source: https://docs.getdeeprecall.com/connect/coding-agents/

> Connect a coding assistant through remote HTTP MCP with OAuth, and add the canonical memory skill.

A coding assistant can use the same brain as your chat assistants when it supports remote HTTP MCP and OAuth.

## Connect through your client

Add a remote MCP server named **Deep Recall** in the client’s MCP settings. Use **HTTP** or **Streamable HTTP** transport and this URL:

```text
https://api.getdeeprecall.com/mcp
```

Complete OAuth in the browser. Do not treat the remote URL as a local executable or put a REST access token in its headers.

For client-specific commands, use the client’s current documentation. Configuration syntax and permission models differ across Claude Code, Codex, Cursor, and other clients; a JSON fragment intended for one is not universal.

## Add the memory skill

[Download the canonical memory skill](https://getdeeprecall.com/skills/deep-recall/SKILL.md) and install it using the client’s supported skill mechanism. The skill teaches when to recall, save, correct, and forget. It does not authenticate the MCP server.

Review how personal memory fits the project. Repository instructions should not automatically expose a person’s private context to teammates or commit recalled personal information into source files.

## Verify tools and attribution

Ask the client to list the available tools, save one harmless preference, and later recall the same memory with its source. Check that you signed into the intended Deep Recall account.

Use the [agent quickstart](/agents/get-started/) and [memory policy](/agents/memory-policy/) for integration behavior. Respect explicit user instructions about whether an assistant may use personal memory in a particular session.