// verdict · checked 2026-08-11 · Replit Agent 3 (docs as of Aug 2026)
Replit Agent
Replit
Builds and deploys the whole thing. Recovers from failure by trying something weirder.
- CAPABILITY
- 5/5 AGENT
- PROOF
- OPERATOR INSPECTABLEYou can run it and read the steps yourself — the record exists for whoever runs it
- TRACE ACCESS
- OPERATOR-OWNEDWhoever runs it holds the record; nothing is published
- LEASH
- Not measuredno execution record scored — measure it yourself
| LOOPSruns until doneyes | Docs: Agent builds, tests its own work in a real browser, fixes what fails and retests before handing back. |
| CHOOSESpicks its own toolsyes | Docs: Agent "figures out the right setup automatically" and decides on its own when to browser-test. |
| ACTSchanges the worldyes | Docs: writes code, "sets up infrastructure," runs the app and clicks through it; checkpoints allow rollback. |
| RECOVERSfixes its own errorsyes | App-testing docs call this Self-Correction: it "automatically fixes any issues that crop up" in its own tests. |
| UNSUPERVISEDnobody watchingyes | Docs: a takeover prompt auto-continues as "Skip" after 10 minutes of no reply, for "Extended Autonomy." |
Capability decisions combine documented behaviour and execution evidence. The proof label on each row shows which of the two decided it, and Leash is reported separately, only when it was measured from an execution record. Primary source: docs.replit.com. Tested against: Replit Agent 3 (docs as of Aug 2026). Our confidence in this dossier: 7/10.
Can you see it work?
the weakest of the five grades aboveOPERATOR INSPECTABLE — You can run it and read the steps yourself — the record exists for whoever runs it. Trace access: OPERATOR-OWNED — Whoever runs it holds the record; nothing is published.
Agent pane expands to show every action taken; work log + checkpoints, but no export.
Where we looked: Replit docs (agent overview, task-lifecycle, app-testing, checkpoints, audit logs), Agent 3 blog, LangChain Replit case study; searched for chat export/trace API — none found. · see the evidence → · re-checked by a second pass built to overturn it
- 2026-08-114/5 AGENT → 5/5 AGENTRECOVERS corrected to yes: Docs and Agent 3 self-test post: browser tests surface failures, Agent then changes the code and continues unattended — not an identical retry.editorial correction — no case was filed
Badges
earned, never for sale
These render from the live dataset, so a badge cannot outlive the verdict it claims. If the verdict changes the badge changes with it, and an unearned one returns 409 rather than an image. No sponsor can buy one, at any price.
Think this is wrong?
bring evidence, not opinionName the criterion and link a trace, a doc, or a recorded run that shows the behaviour. A verdict changed by evidence is the best thing that can happen to this site, and the change goes on the public record with your reason attached. Vendors are welcome, and a vendor's own filing is labelled as such rather than buried.