Registry / verified 2026-09-25

Integrations

Agents earn their keep inside the systems you already run. This is the map of what each cataloged agent connects to, what it does there, and what to check before you hand over a token.

40
Targets
222
Agent connections
6
Kinds

01

Dev tools

Repositories, editors, issue trackers, and tool servers that agents read from and write to.

02

Messaging & comms

Chat, mail, calendar, and phone channels where agents listen and reply.

03

CRM & business

CRM and business systems where agents look up and act on records.

04

Cloud & ops

Infrastructure and operations surfaces agents observe or change.

05

Browser automation

Browser control layers agents drive when no API exists.

06

Payments

Payment rails agents can charge or get paid through.

FAQ

Frequently asked

What counts as an integration?

A documented, named connection between an agent and an external system: an API target, host application, messaging channel, or infrastructure layer. "Works through an API" alone doesn't count. There has to be a first-class connector or documented support.

Why do integrations matter more than feature lists?

Agents earn their keep inside existing systems. An SDR agent without Salesforce, an ops agent without PagerDuty, or a voice agent without telephony is a demo, not a deployment. The integration map is the fastest predictor of whether an agent fits your stack.

Is MCP support the same as an integration?

Not quite. An agent that speaks MCP can attach any MCP server, so it can reach many systems without a native connector. We list MCP support separately and only count a named target when the agent documents support for it directly.

Why is part of each integration's agent list gated?

The full integration matrix is the part of the dataset that took the most verification work. The public view shows the top 80% of agents per integration; the complete tail is available to signed-in accounts.