deepx-codevsphi
Generated from the two spec rows. Green marks the better value where a spec has a clear direction. Everything else is just different.
deepx-code
itmisx · Terminal agent
#101OSSMCP
Panel
7.01 spec wins
- Reliability
- 6.8
- Usefulness
- 6.8
- Cost
- 8.5
- Longevity
- 6.0
“It bundles offline OCR, so the agent can read the screenshot of the error you refused to copy and paste.”
phi
Pulse AI Club · Terminal agent
#28OSSMCP
Panel
6.92 spec wins
- Reliability
- 7.2
- Usefulness
- 6.5
- Cost
- 8.0
- Longevity
- 6.0
“Its extensions talk over a binary protocol on standard input and output, because JSON was apparently the slow part of asking a model.”
Spec by spec
| Spec | deepx-code | phi |
|---|---|---|
| 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 | Yes | Yes |
| Multi-file edits | Yes | Yes |
| Git operations | No | No |
| Browser control | No | NoExternal HTTP fetch is available only through an MCP server you configure. |
| Sandboxed execution | Yes/sandbox defaults to native OS isolation (macOS Seatbelt, Linux bubblewrap) with docker container isolation available as an alternative. | No |
| Multi-agent orchestration | YesIndependent subtasks are split into a plan DAG and dispatched to concurrent subagents, and JS workflow scripts fan work out across agents. | Yes |
| Headless / CI mode | No | Yes"phi run -p" executes one agent loop headlessly, and a headless-strict mode is one of the configurable permission modes. |
| Models | ||
| Backbone | DeepSeek, Xiaomi MiMo, Kimi, Qwen, any OpenAI-compatible endpoint | OpenAI-compatible, Anthropic |
| Bring your own model | Yes | Yes |
| Local models | YesAny custom OpenAI-compatible model can be configured, which covers a locally served one; no specific local runtime is named. | YesAny OpenAI-compatible base URL can be configured, including a locally served endpoint. |
| 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 | MIT |
| GitHub stars | 393 | 524 |
Which one would each critic pick
| Critic | deepx-code | phi | Pick |
|---|---|---|---|
| El Juez | — | — | not enough reviews |
| El Amigo | 7.3 | 7.3 | no preference |
| El Crítico | 6.8 | 6.3 | deepx-code — The code graph resolves symbols through go/types, so callers and impact analysis are exact for Go and fall back to ordinary text tools in every other language. |
| El Profesor | 7.0 | 7.5 | phi — Sub-agents run isolated jobs whose turns stay out of the parent's context while remaining visible in the interface, which separates what the operator sees from what the model reads. |
| La Inversora | 6.3 | 5.5 | deepx-code — 384 stars and a product shaped around the provider it optimises for: a single-vendor bet with no company behind it and no hedge if that vendor's pricing moves. |
| La Jefa | 7.0 | 6.3 | deepx-code — It isolates by default through macOS Seatbelt and Linux bubblewrap rather than asking the user to opt in, which is the first default on this shelf I did not have to argue for. |
| El Hacker | 8.0 | 8.8 | phi — MIT, any compatible base URL including one I serve myself, and tool servers listed by name only until the agent calls list, inspect and invoke to reach them on demand. |
Picks are derived from each critic's own scores. Humans vote on matchups on the duels page.