Yekar.AI
Login

Yekar.AI for platforms

Ship agents inside your product.

Your customers get agent features under your brand. Yekar.AI runs the session underneath - tool access, approvals, limits, and the record of every action. Your customers never have to meet Yekar.AI.

Request accessRead the agent API docs

The session boundary

Every job starts in its own session.

One published agent can serve work arriving from your product. Each API call starts a distinct session with its own message, optional caller credentials, approvals, and action history.

  • Credentials arrive with the call.Pass only the customer credentials the session needs for its bound tools.
  • One job, one session.Messages, tool calls, approvals, and status stay on that session's record.
  • Approvals stay explicit.Sensitive writes suspend the turn until your product submits a decision.
One published agentONE PUBLISHED VERSIONSession AINPUTCREDENTIALSRECORDSession BINPUTCREDENTIALSRECORDSession CINPUTCREDENTIALSRECORD

Wiring it up

An API call, not a rebuild.

Publish the agent, mint a personal API key, and start a session from your backend when work arrives. The same approvals, limits, and action history apply.

  • One personal key for your backend.The key acts with its owner's permissions and can be revoked at any time.
  • Customer credentials only when needed.Send credentials for the agent's caller-bound tools with that session request.
  • Stream the result live.Read messages, tool activity, suspension, and completion over the session event stream.
POST /api/v1/agents/{agentId}/sessions

Authorization: Bearer yk_...
Content-Type: application/json

{
  "message": "Triage support ticket T-4471",
  "callerCredentials": {
    "integrations": {
      "yekar.slack": { "token": "xoxp-..." }
    }
  }
}

-> 202 { "sessionId": "...", "executionId": "..." }

Simplified. The fields are the real ones.Full API reference ↗

What comes built in

The harness is the point.

  • A separate session record

    Every call gets its own messages, tool activity, status, approvals, and durable action history.

  • Approvals & limits

    Sensitive writes wait for a named person. Spend and concurrency stop at the ceiling you set.

  • A record of every action

    Every tool call, its input, its result, and who authorised it - readable long after the run.

  • Versioned releases

    Publish a version; every run is pinned to the one that was live when it started.

  • Usage attributed to the agent

    Every model call is priced when it runs and attributed to the agent and domain that produced it.

  • Your brand in front

    Your customers work inside your product. Yekar.AI stays underneath.

Who builds on Yekar.AI

Built for teams shipping agents to their customers.

Vertical SaaS

Add agent features to the product your customers already run their business on - without hiring an agents team.

Agencies & operators

Run the same published agent across client jobs, passing the right credentials and keeping a separate session record each time.

Start with the job

Ship an agent feature without rebuilding the harness.

Bring the product, the job, and the tools it needs. We’ll show you how the live agent API fits.

Talk to usRead the API docs