CodeAltavskimchi
Generated from the two spec rows. Green marks the better value where a spec has a clear direction. Everything else is just different.
CodeAlta
Alexandre Mutel · Terminal agent
#114OSSMCP
Panel
6.31 spec wins
- Reliability
- 5.8
- Usefulness
- 6.2
- Cost
- 7.5
- Longevity
- 5.7
“Attach files with an at sign and GitHub issues with a hash, so your prompt now has the same syntax as a social media post.”
kimchi
kimchi · Terminal agent
#75OSSMCP
Panel
6.23 spec wins
- Reliability
- 6.2
- Usefulness
- 6.8
- Cost
- 5.8
- Longevity
- 5.8
“It divides work between an orchestrator, a builder and an explorer, which is more organisational structure than some of its users have.”
Spec by spec
| Spec | CodeAlta | kimchi |
|---|---|---|
| Architecture | ||
| Category | Terminal agent | Terminal agent |
| Runs | local | local |
| Platforms | macos, linux, windows | macos, linux, windows |
| Context window | not documented | not documented |
| Protocols | ||
| MCP client | Yes | Yes |
| MCP server | No | No |
| Capabilities | ||
| Runs terminal commands | YesHost tools execute the agent's tool calls, and an in-session "alta" live tool automates local workflows. | Yes |
| Multi-file edits | Yes | Yes |
| Git operations | No | Yes |
| Browser control | No | No |
| Sandboxed execution | No | No |
| Multi-agent orchestration | YesDelegated work and plugin orchestration adapters run through the same session runtime; different sessions execute concurrently. | YesMulti-model mode assigns separate models to orchestrator, builder and explorer roles and delegates tasks between them. |
| Headless / CI mode | No | YesThe repository ships a Terminal-Bench adapter that aggregates usage across session files, which implies unattended runs. |
| Models | ||
| Backbone | Codex, GitHub Copilot, xAI Grok, OpenAI, Azure OpenAI, Alibaba, Anthropic, Gemini, Vertex | kimchi-dev models, Anthropic |
| Bring your own model | Yes | Yes |
| Local models | No | No |
| Cost | ||
| Pricing model | byok | usage |
| Starts at | $0/mo | n/a |
| Free tier | Yes | No |
| Bring your own key | Yes | Yes |
| Openness | ||
| Open source | Yes | Yes |
| License | BSD-2-Clause | Apache-2.0 |
| GitHub stars | 227 | 2,231 |
Which one would each critic pick
| Critic | CodeAlta | kimchi | Pick |
|---|---|---|---|
| El Juez | — | — | not enough reviews |
| El Amigo | 6.0 | 6.5 | kimchi — Pick it if you want role-splitting without wiring it yourself; pick a single-model agent if you would rather not debug three models at once. |
| El Crítico | 5.5 | 6.0 | kimchi — In multi-model mode an orchestrator delegates to role-specific models, so a bad result has three possible authors and nothing documented attributes it. |
| El Profesor | 7.0 | 6.5 | CodeAlta — A session actor per conversation serialises prompts, compacts long histories and journals every step, which places conversation state in the runtime rather than the provider. |
| La Inversora | 5.8 | 6.0 | kimchi — 2,220 stars, a permissive CLI and an API key in the middle: the code is the marketing and the inference is the company. |
| La Jefa | 6.0 | 5.0 | CodeAlta — Zero licence cost, and it installs as a global tool through the package manager my Windows fleet already runs, against Copilot and Codex subscriptions finance already pays. |
| El Hacker | 7.5 | 7.0 | CodeAlta — BSD-2-Clause, MCP servers wired from a documented config, plus local plugins and skill folders that follow the Agent Skills layout so my existing ones drop straight in. |
Picks are derived from each critic's own scores. Humans vote on matchups on the duels page.