Leaderboard / #2 of 13 / Installed or launched
T3 Code
Multi-harness power users, especially anyone mixing Claude and Codex subscriptions or driving agents from a phone.
Weighted feature assessment90.0range 90.0 to 90.0doc coverage 100% / Launchedmeasured: Server probe
Scores by category
Rubric breakdown
Read the scores as a feature assessment. Orca was run end-to-end with stub agents; T3 Code was not run through an equivalent session, so Orca 93 vs T3 Code 90 is a difference in assessed features and evidence, not an experimentally proven gap.
- AAgent breadthweight 155
- BIsolation and parallelismweight 154
- CReview, merge and PR flowweight 155
- DRemote, cloud and mobile reachweight 105
- EOpenness and licenseweight 105
- FMaturity and momentumweight 154
- GLinux setup friction (measured)weight 104
- HPricing and BYO modelweight 55
- IDocs and transparencyweight 53
Verdict
Strengths and weaknesses
Strengths
- Event-sourced orchestrator, durable queue, restart recovery (v0.1.0)
- delegate_task across providers, plus an MCP that lets agents manage threads
- One-click PR with draft and stacked PRs; web and mobile
Weaknesses
- We could not run a session here: it controls agents that are installed and authenticated, and none were
- 2.9k open issues+PRs; v0.1.0 is days old
- Users on X report the sidebar gets noisy at scale
Evidence
What we saw and where it came from
- Hands-on (this box)
- Server ran headless on Linux (ready 1.0 s, 394 MB idle). No agent session run (no authenticated agent CLI usable).
- Server probe
- Headless server port open in 1.01 s, 394 MB idle.
- Known issues
- Public X thread (Oct 2026) of a user finding sidebar hard to use with hundreds of settled threads; 2.9k open issues+PRs on GitHub.
- Notes
- v0.1.0 (Oct 15 2026) made Orchestrator V2 stable: event-sourced command log, durable queue, restart recovery.
Sources
Feature facts
From docs, repos and our probes
- Platforms
- Desktop (Electron: macOS dmg, Windows exe, Linux AppImage/deb), web app, iOS + Android app, headless server binary (linux/mac/win)
- Supported agents
- Claude Code, Codex, Cursor (via Cursor SDK), Grok Build, OpenCode (incl. OpenCode 2), Google Antigravity, Pi, Muse Code, any ACP Registry agent (Devin, Cline, Kimi, Droid...)
- Parallelism
- Threads per agent; each thread writes to its own branch; MCP tool to hand work off to a worktree; delegate_task child agents across providers
- Isolation
- Git branch/worktree per thread (vendor docs). Not containers.
- Review, merge, PR
- Inline diff review, one-button commit/push/PR with generated title/body, draft + stacked PRs
- Remote, cloud, mobile
- Strong: web, mobile, remote environments over authenticated RPC, Tailscale Serve flag, background service
- MCP
- Yes: T3 Code MCP (agents create/launch/message/wait/interrupt threads)
- Cost visibility
- Partial: usage chart in mobile, 'Limited' state + auto-resume when provider limits reset
- Model choice
- Per provider; switch provider/model mid-thread (with caveats)
- Setup
- Download app, or single headless binary: `t3 serve`
"unknown" means we found no source and did not guess.