Put it to work

Use Yekar.AI from Claude, ChatGPT and your coding tools

Ask your agents, follow their conversations and build new ones from the assistants your team already uses.

Captured from the product Demo workspace

Connect Claude, ChatGPT or any MCP client and work with your agents without leaving it.

What this changes for your team.

Yekar.AI serves a Model Context Protocol endpoint. In the Claude or ChatGPT app you add it as a custom connector and sign in to your workspace; in Claude Code, Cursor, VS Code and the Copilot and Codex CLIs you can use a personal API key instead. Your published agents appear as tools, and so does everything the public API does: start and continue conversations, approve a waiting action, create and publish agents, manage their triggers, write knowledge, and fire triggers. With full access allowed, the rest of the product is reachable too - flows and runs, the approvals inbox, connections, domains, knowledge and settings. Every call runs as the person who connected it, with their permissions, and the same approvals and audit trail apply as anywhere else in Yekar.AI. Creating keys, entering secrets, changing who has access and payments stay in Yekar.AI.

How it works in practice.

01
Add the server URL as a connector in Claude or ChatGPT and sign in - or add it with a personal API key in a developer tool.
02
Choose what the app may do on the consent page: ask agents, run conversations, build agents, fire triggers.
03
Work from the assistant, and open the same conversations in Yekar.AI afterwards. Disconnect the app at any time under Triggers → API keys.

What you can plan around.

What to expect, including limits that matter to your team.

The tools are generated from the public API's route definitions, so each one takes the same inputs and runs the same checks as its API route.

Sign-in is standard OAuth with dynamic client registration and PKCE; access tokens last an hour, refresh tokens rotate on every use, and a reused refresh token disconnects the app.

Nothing new is authorised: each call runs through the product's own route and checks, as its owner; admins can see and disconnect members' apps, and signing out of all devices disconnects them.

A conversation that has not finished within the wait budget answers with the agent's state and its conversation id, not a protocol error, so the assistant can report progress and continue.

Complete the operating path

View all capabilities

Put it to workStart agents from connected events

Start work automatically when something happens in a connected app.

Put it to workStart agents through an API trigger

Make a published agent part of your own application or service.

Run it reliablyNever lose track of accepted work

Give every accepted unit of work a durable identity and history before execution begins.

Bring one real process