Set up Cursor Cloud dev environment

[burtenshaw]

Summary

Sets up the development environment for the OpenEnv monorepo and documents cloud-specific setup for future agents.

• Adds AGENTS.md with a ## Cursor Cloud specific instructions section covering project shape, dependency refresh, lint/test commands, how to run an environment server end-to-end, and the optional Gradio web-UI gotcha.
• Configures the startup update script to install uv (if missing) and run uv sync --all-extras.

No application code was modified.

Environment verification

• uv sync --all-extras installs core + CLI + provider extras into .venv/.
• Lint: ruff format --check and ruff check pass; usort check flags two pre-existing files only.
• Tests: PYTHONPATH=src:envs uv run pytest tests/ -q → 1379 passed, 129 skipped (optional env deps).
• Ran the echo_env reference server and completed a hello-world client flow (reset → list_tools → step(echo_message)), echoing back the input message.

Hello-world client log:
echo_hello_world.log

  

[bot-ci-comment]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[cursor]

Alignment Review Report

This PR adds a single new Markdown file, AGENTS.md (+29 lines), documenting Cursor Cloud dev-environment specifics on top of CLAUDE.md. No code, config, CI, or lockfiles are touched.

Automated Checks

• Lint: N/A (PASS) — .claude/hooks/lint.sh only lints Python (src/, tests/, envs/); this diff is Markdown-only. The hook did fail in the review VM, but solely because uv isn't installed there (environmental, not introduced by this PR). Nothing Python changed; defer the canonical lint run to CI.
• Debug code: CLEAN — .claude/hooks/check-debug.sh reports only pre-existing prints/TODOs in src/; zero findings in AGENTS.md.

Open RFCs Context

Current RFCs are In Review (000 project-phases, 001 abstractions, 002 env-spec, 003 mcp-support, 005 agentic-harnesses) or Draft (010 echo-env-token-world-model), plus 004 (rewards/rubrics). None of them cover contributor tooling or dev-environment onboarding docs, so none are relevant to this change.

Tier 1: Fixes Required

None.

Tier 2: Alignment Discussion

Principle Conflicts

None identified. AGENTS.md is contributor-facing onboarding documentation. It does not alter any API surface, reward computation, the client-server boundary, or the MCP/WebSocket dual-API model, so no principle in PRINCIPLES.md or invariant in INVARIANTS.md is affected.

RFC Conflicts

None identified.

Summary

• 0 mechanical issues to fix
• 0 alignment points for human review
• 0 RFC conflicts to discuss

---

Doc accuracy check (for a docs PR, the review is verifying claims against the codebase — all confirmed):

• "~35" environments → exactly 35 dirs under envs/ ✓
• envs/echo_env/server/app.py exists, so python -m echo_env.server.app and uvicorn echo_env.server.app:app are valid ✓
• openenv serve is a print-only stub (src/openenv/cli/commands/serve.py) ✓
• ENABLE_WEB_INTERFACE and CallToolAction are real ✓
• Both named pre-existing usort files (tests/envs/test_grid_world.py, tests/envs/test_julia_env.py) exist ✓
• The Gradio Playground arguments-as-plain-textbox bug is real: gradio_ui.py:182 falls back to gr.Textbox for dict-typed fields ✓

Non-blocking nit: a few hardcoded specifics (~80s suite runtime, ~129 skipped tests, the two named usort files) are accurate today but will drift as the repo evolves. Consider softening them so the doc stays correct without upkeep. Not a blocker — this is a clean, well-verified documentation addition.

  

_{Sent by Cursor Automation: Untitled}

[cursor]

Non-blocking nit: these hardcoded specifics (~80s runtime, ~129 skipped tests) — and the two named usort files on the next line — are accurate right now but tend to drift as the suite grows and lint gets fixed. Consider softening (e.g. "a handful of optional-dep env tests skip") so the doc stays correct without ongoing upkeep.