# Developer Center

---

**Last updated:** October 5, 2026

---

Connect your AI app to FakeParrot so it works from your company's files: the
shared wiki, your private notes and your workflows. Every connection acts as
you, subject to its scopes, current resource permissions and token restrictions.

## Agent bootstrap

1. Identify the execution environment. A local Mac shell can use the folder
   and CLI; Linux shells use the portable CLI with an injected employee secret.
2. On a supported shell, run `fakeparrot auth status`. Otherwise use an existing REST
   token or authenticated MCP connection. Set up sign-in once using the guides.
3. Read a known wiki page from the connected folder. For an unknown page,
   search through the CLI or MCP instead of scanning the whole folder. Use
   live calls for workflows, reports and connections.
4. Discover only what the task needs: `fakeparrot describe QUERY`, the live
   OpenAPI request schema, or MCP `tools/list`. Discovery does not grant access.
5. Load the relevant [page-edit or workflow-run recipe](/developers/api#agent-recipes).
   Keep returned IDs and receipts so a disconnect can be reconciled.

## Pick how to connect

There are three ways in. Use the first one your AI app can reach.

| Way in | What it is | Works on |
|---|---|---|
| [**Mac folder**](/developers/mac-folder) | The FakeParrot Mac app shows your company's files in Finder, as `.md` files any AI app can read and edit. Saved changes sync back to FakeParrot. | macOS 13 or later |
| [**CLI**](/developers/cli) | The `fakeparrot` command lets an AI app with a terminal search the wiki and run and check workflows. The Mac app installs it and signs it in; Linux uses a mounted employee secret. | macOS 13 or later; Linux x86-64/ARM64 |
| **MCP server** | A remote server at `https://fakeparrot.com/mcp` for any app that supports MCP, the open protocol AI apps use to connect to tools. | Any computer, and web apps |

The folder is quickest for reading and editing pages. The CLI reaches live
company data, such as workflow runs and connections. MCP works where the other
two cannot, for example in a web browser without a terminal.

## Which way your AI app can use

| AI app | Mac folder | CLI | MCP server |
|---|---|---|---|
| Claude Code, on your Mac | Yes | Yes | Yes |
| Codex, on your Mac | Yes | Yes | Yes |
| Cursor | Yes | Yes | Yes |
| Claude Cowork | Yes, as a connected folder | Linux terminal with secret injection | Yes |
| Claude desktop app | Only with file tools turned on | No | Yes |
| claude.ai and ChatGPT | No | No | Yes |
| Your own code | Yes, on a Mac | Yes, on Mac or Linux | Yes |

Cowork runs its commands in a separate sandbox. A Linux terminal can use the portable
CLI with an injected employee secret; it cannot use the Mac binary or Keychain.
Use whichever supported surface the session can reach.

## Get started

### Mac folder and CLI

1. [Download the FakeParrot Mac app](/download) and open the installer.
2. Sign in with your work email.
3. Your files appear in Finder under **Locations**, and at
   `~/Library/CloudStorage/FakeParrot-<company>`.
4. The installer also adds the `fakeparrot` command, and a FakeParrot skill
   (a short set of instructions) for whichever of Claude Code, Codex and Cursor
   you have. Run `fakeparrot auth status` in a terminal to check you are signed
   in. The [CLI guide](/developers/cli) lists the commands, and how to install
   the command without the app.

The [Mac folder guide](/developers/mac-folder) explains what is in each
top-level folder, what syncs and how fast, and how an AI app should use it.

### MCP server

Add `https://fakeparrot.com/mcp` to your app and sign in when it asks. The
[MCP server guide](/developers/mcp) has the steps for each app, plus sign-in
with OAuth or an API key for your own scripts. To start a workflow from Zapier or
Make, see [Zapier and Make](/developers/zapier-and-make).

### REST API

Your own code can call the same REST API the web app uses. The
[API reference](/developers/api) covers the base URL, the OpenAPI document,
sign-in, scopes, errors and rate limits.

## Make FakeParrot your AI app's memory

Once connected, an AI app can save memory pages in your FakeParrot Private
folder. Instructions encourage it to retrieve those pages; they do not switch
off the app's built-in memory or guarantee retrieval. The [memory guide](/developers/memory) shows how for
each app, and how to move in what it already remembers.

## Help

Email [support@fakeparrot.com](mailto:support@fakeparrot.com). Report security
issues to [security@fakeparrot.com](mailto:security@fakeparrot.com).