Skip to main content
Glama
TIMI CC MCP Patch Worker

TIMI CC MCP Patch Worker

TIMI CC MCP Patch Worker

一个可本地安装、通过 STDIO 运行的 MCP 服务。它把调用方明确提供的代码上下文发送到 TIMI CC 的 OpenAI-compatible Responses 接口,并返回经过本地路径校验的 Git unified diff;它不会扫描或修改你的仓库。

非 TIMI CC 官方项目。外部模型调用会发送数据并可能产生费用,请先阅读 SECURITY.md。

功能

  • 同步生成或审查候选补丁:generate_patch、review_patch

  • 后台长任务:submit_patch_job、get_patch_job、get_patch_result、cancel_patch_job

  • 本地能力查询:get_capabilities(不调用 API)

  • Responses SSE 流式接收,避免长请求期间完全无活动

  • 严格限制补丁只能修改 allowed_paths 中的路径

  • 拒绝二进制补丁、越权路径和超限响应

  • 只持久化任务摘要、哈希、聚合进度和已校验结果,不持久化完整提示词、文件上下文或推理内容

Related MCP server: DeepSeek MCP Patch Worker

安装

需要 Python 3.10 或更高版本。

git clone https://github.com/prelearn-code/timicc-mcp.git
cd timicc-mcp
python -m venv .venv

Linux/macOS:

.venv/bin/python -m pip install -U pip
.venv/bin/python -m pip install -e .
export TIMICC_API_KEY="your-key"

Windows PowerShell:

.venv\Scripts\python.exe -m pip install -U pip
.venv\Scripts\python.exe -m pip install -e .
$env:TIMICC_API_KEY = "your-key"

不要把真实密钥写进 Git、TOML 示例或命令历史。推荐通过操作系统的密钥管理或在启动 MCP 客户端前设置环境变量。

接入 Codex

将 config.example.toml 复制到 ~/.codex/config.toml,并替换 Python 的绝对路径。Windows 的 command 应改为类似:

command = "C:/absolute/path/timicc-mcp/.venv/Scripts/python.exe"

重启 Codex 后运行 codex mcp list,或在 Codex 中输入 /mcp。示例使用 env_vars 转发已存在的环境变量,并将工具审批设为 prompt,因为每次模型调用都可能发送代码并产生费用。

其他支持 STDIO 的 MCP 客户端可用相同入口启动:

/absolute/path/.venv/bin/python -m timicc_worker.server

配置

环境变量

用途

默认值

TIMICC_API_KEY

API 密钥(也兼容 OPENAI_API_KEY)

无

TIMICC_BASE_URL

Responses API 地址

https://timicc.com

TIMICC_MODELS

额外允许的模型 ID,逗号分隔

无

TIMICC_STATE_DIR

后台任务数据库目录

操作系统的用户 state 目录

TIMICC_JOB_WORKERS

后台并发数

2

默认允许 gpt-5.6-sol、gpt-5.5 和 gpt-5.4。如果兼容网关提供其他 gpt-* 模型,可通过 TIMICC_MODELS 显式添加。

开发与验证

测试不会调用真实 API:

.venv/bin/python -m pip install -e ".[dev]"
.venv/bin/python -m pytest -q
.venv/bin/python -m build

GitHub Actions 会在 Linux、Windows 和 macOS 的 Python 3.10/3.13 上运行相同验证。

安全模型

返回的补丁始终是不可信候选内容。调用方仍需审查完整 diff、应用补丁并运行格式化、lint、构建和测试。allowed_paths 只能约束输出补丁路径,不能识别输入中是否包含敏感数据。

Available Tools

7 tools
cancel_patch_jobB

Request cancellation of a background patch job.

Inputs: job_id (required).

generate_patchC

Generate a validated candidate unified diff without editing the repository.

Inputs: task (required), file_context (required), allowed_paths (required), model, +2 more.

get_capabilitiesA

Return local worker capabilities without calling the external model.

Inputs: no named parameters.

get_patch_jobC

Get the current status of a background patch job.

Inputs: job_id (required).

get_patch_resultB

Return a completed validated patch job result, or its current status.

Inputs: job_id (required).

review_patchA

Review an untrusted candidate patch and return a corrected replacement diff, or an empty patch when no correction is justified.

Inputs: task (required), file_context (required), allowed_paths (required), candidate_patch (required), +3 more.

submit_patch_jobC

Submit a long-running patch generation job and return immediately.

Inputs: task (required), file_context (required), allowed_paths (required), model, +3 more.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    B3.4/5.0

    Scored across 7 tools

    Disambiguation4/5

    Most tools are clearly distinct: generate_patch and review_patch handle direct patch operations, while submit/get/get/cancel_patch_job manage background jobs. Some overlap exists between get_patch_job and get_patch_result, but descriptions clarify that one returns status and the other returns the final result, so confusion is unlikely.

    Naming Consistency5/5

    All tool names follow a consistent verb_noun pattern using snake_case. Verbs (generate, review, submit, get, cancel) are descriptive and uniformly formatted, and nouns (patch, patch_job, patch_result, capabilities) are logical and predictable.

    Tool Count5/5

    With 7 tools, the server is well-scoped for its purpose. It covers both synchronous patch generation/review and asynchronous job management without unnecessary bloat, and each tool fits within a clear workflow.

    Completeness4/5

    The tool surface covers the core patch lifecycle: generate, review, submit asynchronous job, retrieve status/result, and cancel. A minor gap is the lack of a list-all-jobs or delete-patch operation, but these are not essential for the primary use case.

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers