I build AI developer tools and agent workflow infrastructure: focused Pi extensions, CueLoop for queue-driven coding-agent work, and operational systems for governance, scraping, and verification.
These are the six projects I keep pinned. They show the current center of gravity: agent workflow tools, browser/model extensions for Pi, one larger coding-agent workflow system, and one governance/control-plane reference implementation.
| Project | What it does | Start here |
|---|---|---|
pi-cursor-sdk |
Cursor SDK provider extension for Pi models | Quick start |
pi-agent-browser-native |
Native browser automation tool for Pi agents | Browser workflow examples |
pi-codex-goal |
Goal tracking and continuation for long Pi sessions | Install and commands |
pi-oracle |
ChatGPT web handoff extension for Pi | Successful-run transcript |
CueLoop |
Local-first queue for auditable coding-agent work | Workflow diagram |
AI Control Plane |
Host-first AI governance reference implementation | Architecture and boundary |
- For focused agent-tooling packages, start with the Pi extension repos above.
- For a larger workflow system, open
CueLoop. - For systems and operator thinking, open
AI Control Plane,Spartan Scraper, orSplunk TUI. - For end-to-end demo surfaces, open
Promo Studio,agent-eval, orRicochet Rush.
pi-cursor-sdk— Cursor SDK provider extension for pi, adding Cursor's agent loop to pipi-agent-browser-native— nativeagent_browsertool for controlled browser automation in Pi.pi-codex-goal— goal tracking and continuation for long Pi sessions.pi-oracle— ChatGPT web handoff for hard or long-running tasks with archives, async jobs, and saved results.pi-edit-session-in-place— re-edit or delete an earlier user message in the current session branch.pi-zai-mcp— Z.ai MCP server bridge for web search, URL/repository reading, and vision workflows in Pi.pi-stash— stash draft messages and restore them later.pi-local-agents-only— strip globalAGENTS.mdandCLAUDE.mdinstructions for selected projects.pi-copy-user-message— copy the most recent user message to the clipboard.
| Project | What it shows | Start here |
|---|---|---|
Promo Studio |
Multi-agent commerce demo with receipts, diffs, transcripts, and validation | What you are seeing |
Ricochet Rush |
3D browser game with Cursor SDK backed AI board generation/design and local fallback play | Demo and QA |
Spartan Scraper |
Web scraping and research workbench in Go | 5-minute demo |
agent-eval |
Transcript-first evaluation across Codex, Claude Code, and Pi sessions | Outputs |
Morphio |
AI-assisted content workflow monorepo with FastAPI, Next.js, and native routines | Visual tour |
Cloop |
Local loop/task and knowledge assistant, separate from CueLoop | Architecture at a glance |
Splunk TUI |
Rust CLI/TUI for Splunk diagnostics and local-first verification | Quick validation |
- Agent workflow infrastructure
- AI developer tools
- Browser/native automation for coding agents
- Auditable execution, receipts, and verification
- Operational CLIs and control-plane patterns
- Show the outcome before setup when a README is meant for public review.
- Include verification commands, local smoke paths, or proof artifacts where practical.
- Make status and limits explicit instead of implying maturity that is not there.
- Prefer tools that leave inspectable files, logs, receipts, or reports behind.
- Portfolio / writing: FitchHub.com
- GitHub: github.com/fitchmultz
- LinkedIn: linkedin.com/in/mitchelfultz
- X: x.com/fitchmultz