Open-source integration layer giving AI agents authenticated access to third-party applications and MCP-compatible tools.
Corsair is an open-source TypeScript integration layer for applications and AI agents. Instead of making an agent manage provider credentials directly, Corsair resolves credentials at call time and exposes integrations through typed SDK/plugin calls and MCP-compatible tooling.12 Its current first-party product surface includes integrations such as Gmail, Slack, Notion, HubSpot, Google Calendar, Stripe, Airtable, Linear, GitHub and Google Sheets, with support for multi-tenant configurations, webhooks, caching and direct SDK use outside agent workflows.1
The differentiating feature is the permission layer. Corsair lets developers set permission modes per integration and require explicit human approval for destructive or sensitive actions before execution. The public repository describes permission requests as database-backed state that the agent cannot itself approve, while credentials are intended to remain inaccessible to the agent.12 The repository is published under Apache License 2.0, so teams can inspect, extend or self-host the implementation rather than depending exclusively on the hosted service.2
Corsair currently offers Hobby at $0/month, Pro at $200/month and Enterprise at custom pricing. Hobby includes unlimited tool calls, up to 50 connections, up to 100,000 webhook events, unlimited managed permissions/auth pages and up to three team members. Pro removes the stated connection, webhook and team-member limits and adds custom consent-screen branding, Slack support and custom integrations built by the Corsair team.13
The SDK can also be self-hosted on a team''s own infrastructure at no Corsair subscription cost, although the underlying third-party APIs, hosting and operational work remain the customer''s responsibility.1 That gives the product a clear low-cost adoption path for technical teams, while the hosted Pro jump to $200/month means buyers should compare the value of managed OAuth/support against maintaining the open-source stack themselves.
The strongest verifiable evidence today is first-party and inspectable rather than independent. Corsair''s public repository exposes the implementation and Apache-2.0 license, and the website documents the intended security model: credentials are resolved internally, destructive actions can be approval-gated, and multi-tenant deployments isolate credentials and permission evaluation.21 Corsair''s privacy policy describes the hosted service as developer infrastructure around managed OAuth and states that the architecture is designed to limit what Corsair itself accesses; its terms place responsibility for customer-side credential safeguards and, when customers use their own Google OAuth credentials, applicable verification obligations on the customer.43
That architecture is promising, but it should not be confused with independent assurance. Independent exact-product reviews, benchmarks and large user-rating samples are currently scarce, so there is no mature third-party corpus supporting claims about production reliability, security effectiveness, developer experience or long-term integration maintenance. Open source improves auditability, but buyers with strict security requirements should still perform their own code, deployment and threat-model review before granting agents access to sensitive systems.
Corsair is most compelling for engineering teams building agentic products that need many third-party connections while retaining control over credentials, permissions and deployment. It is particularly relevant when multi-tenancy, human approval for sensitive actions, MCP compatibility and the ability to add or modify integrations in source code matter more than a fully managed no-code experience.12
It is less suitable for non-technical teams that want a turnkey automation builder with a large independent review base. It also asks adopters to understand an important architectural choice: the self-hosted path gives more control but shifts database, encryption-key, availability and upgrade responsibilities onto the customer, while the hosted service simplifies parts of OAuth operations without eliminating the customer''s own security obligations.34
Corsair''s clearest strengths are openness, extensibility and security-oriented control points. The combination of typed integrations, OAuth/token handling, MCP support, multi-tenant isolation and explicit approval gates addresses several practical problems that appear when agents move from reading data to taking actions.12 The free self-hosted route and $0 Hobby tier also make serious evaluation possible without an upfront subscription.1
The main weakness is evidence maturity. The feature set and security model are well documented by the vendor and visible in code, but independent reviews, benchmarks and large user-rating samples for the exact product are not yet widely available. That makes Corsair easier to assess architecturally than operationally: teams can inspect what it is designed to do, but should not infer proven production reliability or security merely from the design claims.
Not separately stated in the source review.
Products matched by shared tasks, audience, workflow, product type and capabilities.