Skip to main content
Integrations · Skyvern
Guild agents can launch Skyvern cloud browser tasks and retrieve their results without leaving the agent workflow. Guild stores the Skyvern API key server-side and injects it into every request, so agent code never handles the secret. Credential policies let you restrict which operations each agent can call, and every request passes through Guild’s credential proxy before reaching Skyvern.

Book a demo

See how teams govern agents that use Skyvern.

Talk to sales

Get access to Guild, then connect Skyvern to your first agent.

What the Skyvern integration exposes

The integration provides two tools under the skyvern_ prefix. skyvern_run_task sends a goal prompt and a target URL to Skyvern to start a cloud browser task. skyvern_get_run returns the status, structured output, screenshot URLs, and failure reason for a completed run. Together they cover the launch-and-retrieve workflow for automated web research and staging environment reproduction.

How it works

1

Open credentials in Guild

Go to Access & setup > Credentials in your Guild account or organization. Click Connect and select skyvern.
2

Paste your Skyvern API key

Enter your key in the API Key field. Guild stores it and sends it as x-api-key on every request; agent code never sees it.
3

Add the tools to your agent

Import SkyvernTools from the integration package and spread it into your agent’s tools object. Use pick() to include only the tools your agent needs.

Why Skyvern access needs governing

Skyvern tasks execute goal prompts against real URLs, which means an agent with unconstrained access could trigger unintended actions on staging or production systems. Guild’s credential proxy holds the API key server-side and checks credential policies before any request leaves Guild. You can write ALLOW and DENY rules scoped to specific operations, agents, or workspaces, so one agent can retrieve run results while another is blocked from starting new tasks. Credentials can be scoped to an account, a single workspace, a specific member, or one installed agent. That scope controls which runs can use the Skyvern key at all. Revoking a grant or disconnecting a credential takes effect immediately without changing agent code.

About the Skyvern integration

Guild agents can use Skyvern to launch cloud browser tasks by providing a goal prompt and a target URL. Agents can then retrieve the status, structured output, screenshot URLs, and failure reasons for any completed run. This enables automated web research and staging environment reproduction workflows. The integration is guildai~skyvern.

Authentication

  • Type: API key
  • Header: Guild sends the key as x-api-key: <key>

Setup

1

Connect Skyvern

In Guild, go to Access & setup > Credentials, click Connect, select skyvern, and paste the key into API Key.
See Credentials for workspace, member, and agent-scoped credentials.

Usage

Selecting specific tools

Tools are named with the skyvern_ prefix. Use pick() to include only what your agent needs.

Tools

Tools in the latest published version of guildai~skyvern:

Frequently asked questions

Guild injects the key as the x-api-key header on each outbound request server-side. Agent code never reads or handles the key directly.
Yes. Write a credential policy with a DENY rule on skyvern_run_task scoped to the agents you want to restrict. The credential proxy enforces the rule before the request reaches Skyvern.
Yes. Connect a second credential scoped to that workspace. Guild resolves credentials in order: workspace credentials take precedence over the account default.
No. The catalog entry for this integration lists no webhook events, so Guild does not receive push notifications from Skyvern. Agents must call skyvern_get_run to poll for results.

Put a governed agent on Skyvern this week

Book a demo

See how teams govern agents that use Skyvern.

Talk to sales

Get access to Guild, then connect Skyvern to your first agent.
Also runs on Guild: Browserbase · Cloudflare · Firecrawl · Kernel Read: Docs: Credential policies · Credentials · Tools in agents
Last modified on November 20, 2026