Zapier

Overview

The coSPEC Zapier integration lets you start agent runs from any Zap. Add a single "Create Run" step — it starts the run and waits automatically until the agent finishes (up to 30 days). No webhook setup, no second Zap. Results flow directly into subsequent steps.

Prerequisites

• coSPEC account with an API key (csk_live_...) — see API Keys
• Zapier account (any plan)

Install

coSPEC is currently a private integration (invite-only). No Zapier marketplace listing yet.

Accept Zapier Invite →

1. Click the button above
2. Click Accept Invite & Build a Zap
3. The coSPEC integration appears in your Zapier account

Connect Your API Key

1. In a Zap, add a coSPEC step
2. Click Sign in
3. Paste your API key (csk_live_...)
4. Zapier validates the key automatically

Where to find your key: Dashboard → API Keys → Create API Key

Actions & Triggers

Create Run

1. You add a "Create Run" step to your Zap
2. The step starts the run and waits automatically
3. When the agent finishes (minutes or hours), the step completes with full results
4. Your Zap continues to the next step with the output fields available

Fire-and-forget mode: Set "Wait for Completion" to false — the step returns immediately with just the run ID. Use this for batch or scheduled runs where you don't need results inline.

Testing in the editor: The Zap editor's "Test" button creates a real run but returns sample output fields (the editor can't wait for callbacks). Map fields using the sample data, then turn on the Zap to see real results.

Run Completed Trigger

Fires instantly when a run finishes. Use this for event-driven workflows (alternative to waiting inline).

• Events: Completed, Failed, Cancelled (multi-select, default: all)
• Use case: separate notification Zap, logging, cross-Zap orchestration

Find Run

Search action — look up a run by ID. Returns the full run object with all output fields.

Output Fields

The run output is parsed into convenience fields for easy mapping. These extract the first output of each type — for example, PR URL is the first pull request. If a run produces multiple outputs, use the raw Outputs JSON field for the full list.

Failed Runs

When a run fails, the Status field is failed. Two additional fields help you diagnose:

• Fail Reason — why the run stopped: max_turns, max_cost, timeout, or error
• Error — error details (message and code) when failReason is error

These fields are available in the Zap step output alongside the other fields above.

Troubleshooting