Snapshot: 2026-07-14. Review monthly, before release, and whenever a provider CLI or local default model changes.
This table is a routing reference for humans and host adapters. It is not an automatic routing oracle. opencode, pi, and cmd are model routers, so their "best path" depends on the user's authenticated local model set.
| Model family / need | Primary Polycli provider path | Secondary path | Notes |
|---|---|---|---|
| Claude Code / Anthropic coding agent | claude |
opencode Anthropic models |
Default polycli ask/review now uses official headless claude -p with plan/no-tools/no-MCP constraints, returning synchronous JSON/stream JSON output. Detached tmux TUI remains available in runtime for explicit/internal callers that need an interactive Claude Code session. |
| Gemini | gemini |
none | Official CLI headless -p, --approval-mode plan, JSON/stream JSON. Keep isolated cwd and disabled extensions/MCP for review. |
| Qwen Code / Qwen Coding Plan | qwen |
opencode Alibaba Coding Plan models |
Official Qwen Code default maxSessionTurns=-1 means do not force ask to one turn. Polycli ask uses a bounded maxSteps=20, approvalMode=plan, and --exclude-tools; review uses the same no-tool stance. SDK canUseTool is a better future path if Polycli moves beyond CLI wrapping. |
| Kimi coding | kimi (local kimi-code 0.23.6) |
opencode Kimi For Coding models |
The -p one-shot runner is non-interactive, so ask uses a plain -p invocation and review is prompt-only (like minimax). Current root help advertises -c / --continue for resume-last. Default model comes from ~/.kimi-code/config.toml. |
| MiniMax text / multimodal | minimax via mmx-cli |
opencode MiniMax Coding Plan models |
Use official mmx text chat --message ... --output json --non-interactive; this replaces mini-agent log scraping. |
| OpenCode Go / Xiaomi MiMo / Alibaba / multi-provider routing | opencode |
pi for Xiaomi MiMo |
Local opencode auth list is the source of truth; ~/.config/opencode/opencode.json can show an empty provider object even when credentials and models exist. Current local OpenCode includes Xiaomi Token Plan, Alibaba Coding Plan, Kimi, MiniMax, Anthropic, and OpenCode Go routes. |
| Xiaomi MiMo-V2.5-Pro | opencode with OpenCode Go/Xiaomi Token Plan |
pi default Xiaomi route |
Screenshot state is consistent with OpenCode using Xiaomi Token Plan / OpenCode Go, not an empty provider. |
| DeepSeek V4 Pro | cmd |
opencode-go/deepseek-v4-pro |
User's current Command Code setup routes cmd to DeepSeek V4 Pro. Keep cmd ask/review in --permission-mode plan; rescue may use broader agent mode. |
| Antigravity coding agent | agy |
none | Ask/rescue use Antigravity's session mode with --dangerously-skip-permissions because the CLI has no enforceable non-interactive plan/read-only mode. /review is intentionally unsupported until the upstream CLI exposes a safe review mode. |
| xAI Grok Build CLI | grok |
none | Uses Grok one-shot JSON/streaming JSON mode. ask uses --always-approve; review composes the one-shot path with --permission-mode plan. |
| Copilot / Codex-backed fallback | copilot |
OpenAI Responses API / Agents SDK for new direct integrations | Keep Copilot provider as a fallback, but Polycli ask/review must not pass allow-all tool/path/url flags. Use restricted --excluded-tools and retain --no-ask-user only for programmatic execution. |
| OpenAI GPT / Codex direct programmatic work | not a Polycli CLI provider today | OpenAI Responses API, Agents SDK | For new stateless direct integrations, official SDK/API is more appropriate than wrapping another CLI. |
For domestic/Chinese LLMs with no competitive native coding CLI (MiniMax, DeepSeek, Zhipu/GLM, StepFun) the path is cc-X: point the existing claude runtime (BYOK) or opencode (OpenAI-compatible models) at the vendor's Anthropic-compatible endpoint via ANTHROPIC_BASE_URL + ANTHROPIC_AUTH_TOKEN + ANTHROPIC_MODEL. This is documented in cc-x-endpoints.md (machine-readable source: cc-x-recipes.json). cc-X is not a polycli provider — it rides the existing runtimes with standard env vars; polycli adds no cc-X adapter.
Note: the MiniMax text / multimodal row above is the polycli minimax provider (stateless mmx-cli text/media call), not the MiniMax cc-X coding path. For MiniMax as a coding agent, use the cc-X recipe, not the minimax provider.
Run the automated review-flag subset first, then the manual provider-path probes:
npm run check:provider-paths
opencode auth list
opencode models --refresh
cmd --version && cmd --help
qwen --approval-mode plan --help
qwen --max-session-turns 1 --help
kimi --help
gemini --help
claude --help
copilot --help
mmx text chat --help
pi --help
agy --help
grok --helpnpm run check:provider-paths currently aliases check-review-cli-drift; it verifies the automatable review hard-constraint flags only. The remaining commands in this section are the manual provider-path review.
Qwen's bare help is intentionally minimal in current releases; use both contextual help probes above instead of treating a bare qwen --help omission as drift. If a CLI is not installed locally, record it as skipped rather than failing the release. If a checked flag disappears, update plugins/polycli/scripts/lib/review.mjs, plugins/polycli/scripts/lib/prompt-runtime.mjs, tests, and this table in the same change.
- Claude Code CLI reference: https://code.claude.com/docs/en/cli-reference
- Claude Code permission modes: https://code.claude.com/docs/en/permission-modes
- Gemini CLI headless / CLI docs: https://google-gemini.github.io/gemini-cli/docs/cli/
- Qwen Code approval mode: https://qwenlm.github.io/qwen-code-docs/en/users/features/approval-mode/
- Qwen Code SDK tool callback: https://qwenlm.github.io/qwen-code-docs/en/developers/sdk-python/
- OpenCode CLI / agents docs: https://opencode.ai/docs/cli/ and https://opencode.ai/docs/agents/
- GitHub Copilot CLI command reference: https://docs.github.com/copilot/reference/cli-command-reference
- GitHub Copilot CLI modes/autopilot: https://docs.github.com/copilot/concepts/agents/copilot-cli/about-copilot-cli
- MiniMax CLI docs: https://platform.minimax.io/docs/token-plan/minimax-cli and https://github.com/MiniMax-AI/cli
- xAI Grok Build CLI docs: https://docs.x.ai/docs/grok-build/introduction
- OpenAI Responses API: https://platform.openai.com/docs/api-reference/responses/create
- cc-X endpoint recipes (no native CLI cluster): ./cc-x-endpoints.md (per-vendor source URLs live in ./cc-x-recipes.json)