v0.1.59 — Your agents show their work
Released 2026-06-25. Multi-platform signed/notarized DMG + EXE + AppImage. Auto-update from v0.1.58 happens on next launch.
Mission Control (v0.1.58) let you dispatch and supervise a whole team of agents. v0.1.59 makes every mission prove what it did.
Artifacts — typed, verifiable, from the real run
Every finished mission now opens a Review drawer of typed artifacts, derived from the actual run — never a summary it made up:
- 📋 Plan — the steps it set out to do
- 🧭 Walkthrough — what actually happened, step by step, pass / fail / skipped
- 📁 Files — every file it touched
- 📄 Output — the final result
- ⚠️ Error — if it stopped, exactly why
This is the “show real work” surface — what lets you trust an agent without watching it the whole time.
Verify in browser — proof-of-work, captured
If a mission built something you can open in a browser, hit ▶ Verify in browser. AgentFoundry drives the embedded browser to actually run the agent’s page and captures a screenshot as proof, attached right to the mission as a verification artifact. The agent doesn’t just claim it works — it shows you.
Both are on the real substrate: real run records, real browser tools, real screenshots. No mocks.
Where this leaves us vs Antigravity
With Mission Control + Artifacts + browser-verify, AgentFoundry now matches agent-first IDEs like Google Antigravity on the surfaces that build trust — and gives you what a single-vendor IDE structurally can’t: multi-model BYOK, an agent harness you own, and a Business mode for the non-coders who actually oversee the work.
Upgrade notes
| Change | Who notices | What to expect |
|---|---|---|
| Mission artifacts | Mission Control users | Done / failed missions get a Review button → the artifact drawer. |
| Verify in browser | Anyone building web output | A Verify button appears when a mission touched an .html file. |
Auto-update from v0.1.58 happens on next launch. No migration steps.