About this benchmark
An independent look at the apps that run several coding agents in parallel with worktrees, sessions and review. Built from docs and feature testing, with the limits stated plainly.
Questions
Who made this?
Arjun Shah ran the benchmark on 2026-10-19. Every number on this site comes from the logs under runs/ and the curated data file in the benchmark repo.
Is this a test of which tool writes the best code?
No. Code quality depends on the model and agent inside the orchestrator. We measured orchestration features and mechanics, and no real-model task runs feed the scores.
Why did real-model runs not happen?
No agent CLI was logged in on the box. A Groq free-tier key was tried through opencode, and 0 of 9 attempts passed because a single request exceeded the per-minute token limit.
Which tool is David Ondrej's?
Cloudroom (cloudroom.dev; davidondrej/cloudroom-core and cloudroom-gui, Apache-2.0). Its weighted feature assessment is 72.0 (range 62.0 to 82.0) with 80% doc coverage. There is a public dispute on X between its maker and T3 Code's author; we cite neither side's claims as evidence.
Why is Orca's star count flagged?
GitHub reports about 97.9k stars, which is unusually high for a 7-month-old repo. We did not audit star provenance, and stars are only one input to the maturity criterion.
What does n/r in the table mean?
No evidence for that criterion, so it is unranked. The score range shows the result if it were 0 or full marks; the point score is the midpoint.
Is Orca really better than T3 Code?
Not proven. Orca was run end-to-end with stub agents; T3 Code was not run through an equivalent session. 93 vs 90 is a difference in assessed features and evidence.
Can I re-weight the scores?
The leaderboard shows four weight profiles. The raw criterion scores are in assets/data.json if you want your own.
Were any vendors paid or contacted?
No. Nothing was bought and no accounts were created.
Named in secondary sources, not checked
BridgeMind (paid, Aug 2026 beta), CodeAgentSwarm (closed, open beta), Morphy (closed; its author wrote the main roundup we used), Nimbalyst (Crystal successor), JetBrains Air, Zed Threads, GitHub Copilot app, Google Antigravity 2.0, Pane, Herdr, Tern, bb, Capy (named in X tier-list replies; unverified).
Everything we cited
- T3 Code site
- T3 Code repo + v0.1.0 notes
- T3 Code architecture
- Synara site/docs
- Synara install
- Synara repo
- Orca repo
- Orca orchestration docs
- David Ondrej GitHub (Cloudroom)
- Cloudroom announcement (LinkedIn)
- David Ondrej on X: Cloudroom GA
- Theo on X re: T3 Code
- Synara maintainer on X
- Conductor pricing
- Claude Squad repo
- Superset
- Emdash
- Paseo
- Vibe Kanban repo
- Sculptor
- Terragon shutdown
- Omnara pricing
- Cursor Agents Window
- Warp orchestration
- Amp pricing
- Secondary: appmorphy roundup (author builds a competitor, disclosed)
- Secondary: Augment Code OSS orchestrators
- Secondary: Nimbalyst roundup (vendor)
Site built as static HTML. Charts and art use DitherKit Lite (the SuperCompress brand-pack port of Tripwire dither-kit) plus a small Bayer renderer for the heatmap, radar and hero.