Coding Agents

If you're the platform or internal-dev engineer rolling out Appstrate inside a company, a coding agent is your fastest path. Install the Appstrate Agent Skill into Claude Code, Cursor, Antigravity, or any AGENTS.md-aware host, and delegate the configuration work (authoring agents, wiring providers, rotating keys, testing impersonation, publishing AFPS packages) to an agent that already knows Appstrate's CLI, REST API, and conventions.

Instead of opening the dashboard, reading the docs, and hand-typing curl or appstrate api, you describe what you want ("wire a Gmail provider and create an agent that summarizes new emails every morning") and the coding agent executes against your active profile.

This page is the why and when. For the how (install, targets, flags, update, customise), see Using Appstrate / Skill.

Why this pattern

• Faster config loops. Authoring an agent + provider + webhook normally takes 15 min of UI clicks or precise API calls. A coding agent does it in one prompt, with the manifest conventions baked in.
• Multi-instance workflows. The skill uses named profiles (dev, staging, prod). Your agent can diff a config between them or replay a prod change on staging first.
• End-user impersonation checks. Ask the agent to validate that the same flow works across 3 end-users. It loops through them with Appstrate-User headers.
• Less context-switching. Ship an internal feature that uses Appstrate from the same chat where you wrote the code. No jump from IDE to dashboard to terminal.
• AFPS packaging. The agent knows the manifest shape, dependencies, semver, and registry flow. It packs and publishes without you looking up the spec.

Use cases that land well

Once the skill is loaded, these prompts work out of the box:

• "Create an agent called email-triage that reads my Gmail inbox and posts a summary to Slack. Wire the Gmail and Slack providers for end-user eu_alice."
• "Rotate the API key named ci-deploy. Revoke the old one only after the new one is verified."
• "Pack my local ./my-skill folder as an AFPS and publish it to my instance under @acme/[email protected]."
• "List every agent that's failing more than 10% of the time this week, show me the last run's stderr."
• "Replay run run_01HZABC against end-user eu_bob to check if the provider keys work for him too."
• "Diff my dev profile's agent list with prod, tell me what's missing on prod."

The agent reads the skill's SKILL.md and references on demand, picks the right CLI command or REST call, runs it, and reports back. If something fails, it stops and shows the stderr instead of retrying blindly.

Reference

• Using Appstrate / Skill: install, targets, flags, update, customisation.
• CLI reference: the 9 commands the skill drives.
• Skills as a feature: how Appstrate itself consumes Agent Skills.
• Compare: how a coding-agent + Appstrate combo contrasts with coding-agent alone.