Agent Maestrovsdev-3.0
Generated from the two spec rows. Green marks the better value where a spec has a clear direction. Everything else is just different.
Agent Maestro
Joouis · Agent harness
OSSMCP
Panel
6.32 spec wins
- Reliability
- 5.8
- Usefulness
- 6.0
- Cost
- 8.3
- Longevity
- 5.0
“Its hosted web-search tool is executed through Exa, so your editor now subcontracts curiosity.”
dev-3.0
h0x91b · Agent harness
OSS
Panel
6.32 spec wins
- Reliability
- 6.0
- Usefulness
- 6.5
- Cost
- 7.7
- Longevity
- 5.2
“Your coding agents are now cards on a Kanban board, which means they will be moved to Done by someone who has not read them.”
Spec by spec
| Spec | Agent Maestro | dev-3.0 |
|---|---|---|
| Architecture | ||
| Category | Agent harness | Agent harness |
| Runs | local | local |
| Platforms | macos, linux, windows | macos, linux, windows |
| Context window | not documented | not documented |
| Protocols | ||
| MCP client | Yes | No |
| MCP server | No | No |
| Capabilities | ||
| Runs terminal commands | YesAgent Maestro is a control plane; command execution and file edits happen inside the Roo Code, Cline or CLI agent it drives. | Yes |
| Multi-file edits | Yes | YesFiles are changed by the agent running in each card's worktree, not by dev-3.0 itself. |
| Git operations | No | YesEvery card is a real git worktree and branch; the agent opens the pull request and enables auto-merge, and the card then shows the PR number and CI state. |
| Browser control | No | No |
| Sandboxed execution | No | No |
| Multi-agent orchestration | YesUp to twenty concurrent Roo Code tasks can run at once. | Yes |
| Headless / CI mode | YesTasks are created and managed entirely through REST APIs, which the README calls headless AI agent control. | No |
| Models | ||
| Backbone | Anthropic, OpenAI, Gemini, GitHub Copilot | Claude Code, Codex, Gemini, Cursor Agent, opencode |
| Bring your own model | Yes | Yes |
| Local models | No | No |
| Cost | ||
| Pricing model | byok | byok |
| Starts at | $0/mo | $0/mo |
| Free tier | Yes | Yes |
| Bring your own key | Yes | Yes |
| Openness | ||
| Open source | Yes | Yes |
| License | MIT | Apache-2.0 |
| GitHub stars | 206 | 301 |
Which one would each critic pick
| Critic | Agent Maestro | dev-3.0 | Pick |
|---|---|---|---|
| El Juez | — | — | not enough reviews |
| El Amigo | 6.8 | 6.5 | Agent Maestro — Pick this if you already run Roo Code or Cline in VS Code and want a script to start their tasks; pick a plain terminal agent if you would rather the editor were gone. |
| El Crítico | 6.0 | 5.8 | Agent Maestro — It controls extensions it does not ship. A Roo Code or Cline release can change the surface underneath it, and twenty concurrent tasks share one editor process. |
| El Profesor | 6.8 | 6.8 | no preference |
| La Inversora | 5.5 | 5.8 | dev-3.0 — Permissively licensed, 252 stars, one maintainer, and the product is a shell around five other companies' agent CLIs. There is no asset here that anyone would buy. |
| La Jefa | 5.0 | 6.0 | dev-3.0 — Nothing to buy across sixty desks, and the one thing that reaches my existing process is a card that shows its pull-request number and live continuous-integration state. |
| El Hacker | 7.8 | 7.3 | Agent Maestro — MIT, and it flips the arrow: Claude Code, Codex or Gemini CLI point at my editor instead of a provider, because it answers /messages and /chat/completions itself. |
Picks are derived from each critic's own scores. Humans vote on matchups on the duels page.