Apps & Agents

FreeRide works in the AI apps you already use. Every app reads and writes the same projects, so an idea you save in Claude chat is there when Claude Code, Codex or Cursor starts work the next day.

Coding agents are where FreeRide does the most: sessions, work logs, docs that stay current. In chat apps it's lighter: catch up on a project, save ideas and decisions.

Three steps, and only one is per repo

Linking a repo tells FreeRide which project a folder belongs to. The setup writes a small file, .freeride.json, with the project's id (no secrets), and adds FreeRide's rules to both CLAUDE.md and AGENTS.md. Commit them. From then on the repo is ready in every coding agent, whichever one you ran setup in, and teammates who clone it are linked too.

You can run setup yourself, or let the agent offer it: when you talk about a FreeRide project in a repo that isn't linked yet, it offers once to link it.

Chat doesn't need linking. A chat has no folder. You name the project ("catch me up on Mealwise"), or Claude asks which one. One chat can touch several projects.

Claude Code

The deepest integration. Install the plugin, then link each repo:

/freeride:setup

Setup asks which project the repo belongs to (or creates a new one), writes .freeride.json, and adds FreeRide's rules to CLAUDE.md and AGENTS.md. It shows every change and asks once before making it.

From then on, every conversation in that repo starts with a briefing on the project. Your agent loads feature context before work, logs what it builds, and keeps the docs current. Check the setup any time with /freeride:status.

Codex

The same plugin works in Codex (the app and the CLI). Install it from Plugins, sign in when asked, then open a repo and ask Codex to "set up FreeRide for this repo". Setup links the repo and adds FreeRide's rules to AGENTS.md (and CLAUDE.md, for Claude Code).

Codex asks you once to approve the plugin's hooks (a "Review hooks" window). Allow them: they're FreeRide's reminders, such as suggesting to log work after a build. FreeRide works without them too, just with fewer reminders.

Codex has a smaller limit on how much a tool can return, so FreeRide sends it a shorter project briefing. The facts are the same; anything left out is one search away.

Cursor

The plugin works in Cursor too. Install it, click Authenticate next to the freeride server on the plugin's page, then ask the agent to "set up FreeRide for this repo". Setup links the repo and adds the rules to AGENTS.md (and CLAUDE.md), including a line that names the project, because Cursor can't tell FreeRide which repo it's in.

Two Cursor habits to know:

• Use Agent mode. Ask mode is read-only, so Cursor blocks FreeRide's tools there.
• Tool approvals. Cursor asks before each tool call. You can let FreeRide's tools run without asking.

The nudges work in Cursor too. Two of them come a step later than in Claude Code, because Cursor has no matching moment: the "FreeRide is connected" reminder right after the first edit instead of before it, and the reminder after compaction at the agent's next tool call.

Claude chat and Cowork

Install the plugin on claude.ai (Customize → Plugins). Then open the plugin's Connectors tab and connect FreeRide. That's the sign-in, and nothing works in chat until it's done.

Then just talk about your projects:

• "Catch me up on Mealwise" → a short card with where the project stands
• "New idea for Mealwise: import recipes from a URL" → saved, with a receipt you can undo
• "What did we decide about pricing?" → searched, not guessed

If Claude isn't sure which project you mean, it asks. It never saves to a guessed project.

Make it automatic. Chat apps don't read CLAUDE.md. For the most reliable experience, add one line to claude.ai Settings → Instructions for Claude:

For my projects, use FreeRide: catch up on the project before we discuss it, and save new ideas and decisions to FreeRide, not to your built-in memory.

Claude's memory and FreeRide. Claude's own memory remembers you: how you like to work, your preferences. FreeRide holds your projects: features, ideas, decisions with their reasons, docs and work history. When you save something to a project, FreeRide may offer once to note in Claude's memory that the project lives in FreeRide, so later "remember this" requests for it land in the right place.

Claude Desktop. If the Claude Desktop app is connected, your claude.ai chats may run their tools through it. FreeRide then uses the Desktop app's sign-in, so connect FreeRide there too (Settings → Connectors). If chat says FreeRide "isn't responding", reconnecting there usually fixes it.

One thing to know. A plugin installed on claude.ai also appears in Claude Code, and so does the chat connection. If FreeRide shows up twice in Claude Code, run /freeride:setup. It offers a one-line setting that turns off the chat copy in Claude Code only; chat keeps working.

ChatGPT

Coming. The same server and cards work in ChatGPT, and we're packaging the plugin for OpenAI's directory.

Moving between apps

Nothing to move. Every app talks to the same FreeRide account:

• Brainstorm in Claude chat, decide something, save it. Tomorrow, Claude Code's begin_work on that feature shows the decision.
• Build in Codex. Ask Claude chat "what was built this week?" and the work is there.

Related

• Quickstart → — install and link your first repo
• How It Works → — what happens during a session
• Troubleshooting & FAQ → — when something doesn't connect