Inbox Dog: Give agents full Gmail access
Inbox Dog is a Gmail OAuth service that lets AI agents read, write and search a mailbox using its CASA Tier 2 verified credentials.
What is Inbox Dog?
Inbox Dog supplies OAuth credentials that let an AI agent work with Gmail beyond read-only access. Because Google gates Gmail write scopes behind a CASA Tier 2 audit, the service has already completed that review and lets builders use its credentials instead of running their own. Agents can send mail, create drafts, manage labels and search across inboxes, with token refresh handled automatically. There are three ways to connect: a no-code flow at inbox.dog/connect, an MCP server for clients such as Claude, Cursor and Windsurf, and an npm package for code that needs raw tokens and a REST API. The project is MIT licensed, so it can be self-hosted on a Cloudflare account, though production Gmail write access then requires your own Google OAuth credentials and CASA audit.
Focus areas: MCP.
Who is Inbox Dog for?
- Developers building AI agents that act on Gmail
- Teams using MCP clients such as Claude, Cursor or Windsurf
- Builders who want to self-host on their own Cloudflare account
How much does Inbox Dog cost?
Inbox Dog starts at Free. Always check the vendor’s own pricing page before you buy.
| Plan | Price | What you get |
|---|---|---|
| Hosted | Free | Unlimited API keys |
Is Inbox Dog worth it?
A practical shortcut for agent builders who need Gmail write scopes without running Google's audit themselves
Where Inbox Dog is strong
- CASA Tier 2 verified credentials cover Gmail write scopes
- Ships as an MCP server plus an npm package and REST API
- Automatic token refresh at no cost
Where it falls short
- Only covers Gmail, not other email providers
- Self-hosting production write access still needs your own OAuth credentials and CASA audit
How do you get started with Inbox Dog?
- 01Create an API key on inbox.dog
- 02Connect Gmail through the no-code page, the MCP server or the npm package
- 03Point your agent at the credentials and start reading, writing and searching
Inbox Dog: frequently asked questions
Do I need my own Google CASA audit?
No. Inbox Dog states it is CASA Tier 2 verified, so you can use its OAuth credentials for Gmail write scopes instead of completing the audit, which it says costs $550 or more and takes weeks to months.
What can an agent actually do with Gmail?
The service covers read, write and search: sending email, creating drafts, managing labels and searching across inboxes, rather than read-only access.
How do I connect it to my agent?
There are three routes: a no-code connect page, an MCP server added as a tool server for clients like Claude, Cursor and Windsurf, or an npm package for code that handles OAuth and uses the REST API.
What does it cost?
The hosted service is listed as free, with unlimited API keys and no card required. The source is MIT licensed and can also be self-hosted on your own Cloudflare account.
Inbox Dog alternatives
The closest tools in our directory, matched on category, focus areas and audience.
| Tool | Best for | Free plan | Starting price | Focus areas |
|---|---|---|---|---|
| Inbox Dog | Agent builders needing Gmail write access without a CASA audit | Yes | Free | MCP |
| Pipedream | Developers adding authenticated API integrations to agents and apps | Yes | $99/mo | MCP |
| RapidMCP | Developers connecting existing REST APIs to MCP-compatible AI agents | Yes | $29/mo | MCP |
| Langflow | Developers prototyping and deploying agentic and RAG workflows | Yes | — | MCP |
| Maxkb | Open-source AI assistant for enterprise with RAG and LLM. | Unknown | — | MCP |
| Chainlit | An open-source Python library for quickly building and sharing user interfaces for your LLM applications. | Yes | Free | — |
| Termdock | Terminal-centric AI dev environment unifying workspaces, terminals, and Git. | Yes | Free | — |
| Composio | Connecting AI assistants and agents to many apps | Yes | $29/mo | — |
| Verdent AI | Solo builders and small teams shipping full products | Yes | $5/mo | — |
| Agno | Teams that want open-source agents running on their own infrastructure | Yes | — | Agent Deployment, MCP |
| HumanLayer | Teams reviewing agent plans and diffs together | Yes | — | — |
