
Sign in to rate this tool.
Overview

Corsair is an open‑source layer that sits between AI agents and the external apps your users rely on. Hosted on GitHub, it lets developers add OAuth‑based integrations, permission checks, multi‑tenant support, and hosted or self‑hosted execution for the agent's backend. The project emphasizes keeping the code that talks to each provider simple and up‑to‑date.
When you start a new instance, you create it with `createCorsair({ multiTenancy: true, database: pool, kek: process.env.C })`. The `pool` drives a shared database that stores users' OAuth tokens and permission states, while the `kek` encrypts those tokens. After that, the dashboard shows a list of available plug‑ins, Gmail, Slack, Notion, Google Calendar, Stripe, Airtable, Linear, GitHub, PostHog, Resend, Sheets, Discord, and Tavily. For each plug‑in the dashboard indicates whether a user is connected, missing credentials, or has incomplete approval. The UI also displays the user who added each integration and shows status updates once a connection succeeds.
You interact with a connected Spark via simple JSON requests. For example, a request to pull all orders that have not been shipped from Airtable would contain a query JSON and the key of the Airtable integration. The agent, once authenticated, fetches the records, filters by the “Shipped” field, and returns the subset in a JSON array. Another use case is sending a calendar invite through Gmail and Google Calendar: the request contains the invite details, the target email address, and the user's Gmail credentials. The agent calls the Gmail send API, then triggers a create‑event request on GCal. Each step logs success or shows an error if the token has expired or the API endpoint has changed.
Permission controls are granular. The default mode is “safe,” allowing non‑destructive read operations. If an action will modify data, such as deleting a Slack thread or updating a Stripe customer, Corsair requires explicit approval. The approval process gives the user a link to confirm the change, preventing silent updates. This is useful for services where accidental destructive actions would cause loss of data.
Corsair is aimed at teams building AI products that need to talk to many third‑party services without writing ad‑hoc integration code for each.
Details
- Pricing
- Free
- Last checked
- 19h ago
- Indexing
- noindex, follow
- Category
- Chat
Suggest an edit
Sign in to suggest an edit.
Pros and cons
Pros
- Rapid integration set up for common business apps
- Built‑in multi‑tenant and permission controls
- Public repository invites community contributions
Cons
- No published pricing or support channels
- Limited to integrations listed on the site
- Requires developer effort to host and maintain the service
- No explicit documentation on scaling limits or API rate handling beyond the listed examples

