Verify changes to the copilot-kit-ag-ui demo by building and running its server. Check API routing, cross-origin access settings, and build configuration, then use real credentials to confirm that a requested chart appears.
Build and drive the copilot-kit-ag-ui demo to verify changes end-to-end.
Review the instructions the agent will receive before you install.
Build and drive the real server. npm run typecheck is CI's job, not evidence.
The server boots and serves everything except a live agent turn with stub IDs:
npm install && npm run build
ANTHROPIC_API_KEY=sk-ant-test ANTHROPIC_ENVIRONMENT_ID=env_x \
ANTHROPIC_AGENT_ID=agent_x ANTHROPIC_AGENT_VERSION=1 PORT=8799 npm start
Probes that exercise the wiring:
GET :8799/ returns the built web/dist/index.html (there is no SPA catch-all, so unknown paths 404).GET :8799/api/copilotkit/info lists the registered agents, confirming agent id and class registration without an API call.POST :8799/api/copilotkit/agent/financial-assistant/run with an AG-UI body ({"threadId":"t1","runId":"r1","messages":[{"id":"m1","role":"user","content":"hi"}],"state":{},"tools":[],"context":[],"forwardedProps":{}}) reaches the AG-UI adapter and fails with a 401 from Anthropic. That 401 in the SSE RUN_ERROR proves the full route-to-SDK path.OPTIONS with an Origin header. With ALLOWED_ORIGINS set, allowed origins get Access-Control-Allow-Origin echoed and others get none.VITE_COPILOT_RUNTIME_URL=... npm run build then grep web/dist/assets for the URL to confirm build-time baking.ant apply agents environments once, npm run dev, open http://localhost:5173, send a prompt that triggers a visual tool ("show me a growth projection for $500/month at 7%") and confirm the chart renders inline.
claude-lock.json (from ant apply agents environments) or the three ANTHROPIC_* ID env vars. Missing both is a deliberate boot failure.web/dist exists, so run npm run build before npm start probes.No reviews yet. Tell others what you thought.
No login needed. Share how it worked for you, ask a question, or note where it did not run.
No comments yet. Be the first to write one.
Guide a first computer-use agent run from environment checks to a browser-only task and a viewer of the run history. Explain each stage so users can understand the output and review screenshots afterward.
Build and run computer-use-demo in Docker, then use Playwright to test its Streamlit interface without opening a visible browser. Save conversation transcripts, API exchanges, and screenshots to inspect behavior after changes.