✨ Yep Agent

Yep Agent is a coding agent built on top of YepCode that turns a prompt into real, runnable YepCode processes (JavaScript or Python). It works directly on your workspace codebase (processes, modules, schemas, README files), and helps you iterate until the automation is ready to run in production.

• In YepCode Cloud: open any process in your team and launch Yep Agent to start prompting against your workspace.
• From the public entry point: use the AI landing at /ai (you can create an account at cloud.yepcode.io).

• Generate a complete process: source code + input parameters schema + README.
• Create reusable modules: extract integrations (SDK clients, helpers) into modules.
• Add dependencies automatically: use any npm/PyPI packages via dependencies.
• Run and validate: trigger test executions to inspect datasources, inspect logs, and then iterate.
• Configure executions on your workspace: like scheduled executions once the process is correct.

Yep Agent is designed to behave like a real developer working in your repo:

1. It analyzes your workspace (existing processes, modules, conventions).
2. It proposes an implementation plan and waits for confirmation before applying changes.
3. It generates code incrementally (often creating modules for external APIs and helpers).
4. It validates by executing inside YepCode, then fixes issues based on results.

Yep Agent runs in a secure sandbox, aligned with YepCode’s execution model. When it needs to interact with real services (databases, third-party APIs), it does so by launching executions in your YepCode team context—where your configured credentials exist—without exposing secret values to the LLM.

For guidance on managing secrets/configuration in your workspace, see Team variables.

Aim for one automation goal, plus constraints and outputs. Example (based on the showcase):

Scan my Supabase buckets, find TXT files, convert them to audio using ElevenLabs, and store them in an audio/ folder. Let me pick the voice/model from my ElevenLabs account, and send a Slack notification with the converted files.

If your process needs credentials (e.g. Supabase, ElevenLabs, Slack), Yep Agent will detect that and either:

• use credentials you already configured in your workspace, or
• ask you to create them before it can run validations.

After the plan is approved, the agent will typically:

• generate modules for integrations (e.g. supabase, elevenlabs)
• add required dependencies
• build a parameters schema so you can choose options in a form (for example voice/model)
• run a test execution to verify behavior end-to-end

When something isn’t right, tell the agent what you observed (for example: “you didn’t find files that exist in the root folder”). It will inspect the implementation, adjust assumptions (e.g., bucket root vs a subfolder), push updated code, and re-run validations.

Once correct, ask for production wiring:

• Scheduling: “Schedule this process every Monday at 9:00 AM.” See scheduled executions.
• User-friendly inputs: “Expose a dropdown with titles for the voices and models.”

• Be explicit about IO: where data comes from, where it goes, and what “done” looks like.
• Name the constraints: folders, formats, naming conventions, rate limits, timeouts.
• Request ergonomics: ask for good parameter schemas, defaults, and helpful descriptions.
• Prefer modules for integrations: it keeps process code small and reusable.

If you want the behind-the-scenes architecture and security design, read the build story: “Yep Agent: The Making Of”.