nim-mcp
Enables interaction with NIM Labs project tracking through a standardized MCP interface, supporting project management tasks like reading, verifying, migrating, and rolling up NIM projects alongside other trackers.
README
nim-mcp
A standardized MCP server for NIM Labs — the 5th tracker in the
tracker-MCP family. It exposes NIM through the same normalized
contract as shotgrid-mcp,
ftrack-mcp, kitsu-mcp and
ayon-mcp, so an agent (or the hub)
can read, verify, migrate and roll up NIM projects alongside the other four.
No nim_core SDK dependency — it drives the NIM HTTP API (/nimAPI.php) directly over REST.
NIM data model
Job (the project) ──► Shows (~sequence/episode tier) ──► Shots ──► Tasks ──► Files/Versions · Elements · Renders · Review items
└─► Assets └─► Tasks
NIM is the only tracker with a fixed three-level hierarchy (Job → Show → Shot); a Job is the
project container, Assets hang off the Job. It's also the strongest reference for bidding / scheduling /
timecards (the job lifecycle literally starts at BIDDING). See COMPARISON.md for the
full 5-way mapping.
Install
pip install -r requirements.txt # fastmcp + requests
cp .env.example .env # then fill in the values
Configure (.env)
NIM_URL=http://your-nim-host # the NIM VM/app host; the API is <NIM_URL>/nimAPI.php
NIM_API_USER=yourbot # a NIM username (or numeric user id)
NIM_API_KEY=... # per-user key (Admin ▸ Security ▸ Options); may be blank if key-optional
Auth is sent as X-NIM-API-USER / X-NIM-API-KEY headers. NIM can run key-optional (the API is open) —
in that case reads work with a blank key; enabling keys is recommended for anything networked.
Register with Claude Code
claude mcp add nim -- python3 /path/to/nim-mcp/server.py
Tools (25)
| Tool | NIM function(s) |
|---|---|
whoami |
testAPI + getUsers — version, key validity, resolved user |
list_jobs |
findJobs (all) / getUserJobs (a user's assigned) |
get |
get{Job,Show,Shot,Asset,Task}Info |
list_shows · list_shots · list_assets |
getShows / getShots / getAssets (parent ID=) |
list_tasks |
getTaskInfo&class=shot|asset&itemID= |
list_task_types · list_statuses · list_users |
getTaskTypes / get*Statuses / getUsers |
new_job · new_show · new_shot · new_asset · new_task |
addJob / addShow / addShot / addAsset / addTask |
update_task · set_task_status |
updateTask (taskStatusID, validated by name) |
add_render · add_element · add_file · list_versions |
addRender / addElement / addFile / getVersions |
upload_review · log_time |
uploadReviewItem / addTimecards |
nim_call |
escape hatch — any q= function + params |
project_summary |
the normalized cross-tracker snapshot (below) |
Every new_* / update_* / set_* tool takes dry_run: "plan" (echo the call, no server contact) or
"preflight" (resolve parents + validate the status against the live list, write nothing).
⚠️ Not in NIM's API (UI-only — these tools can't exist)
NIM's production-business layer has no REST API — each function below returns Invalid query, so
nim-mcp deliberately does not wrap them (verified live against NIM 7.2.9):
| Domain | Status |
|---|---|
| Bids (+ line items, templates) | ❌ UI-only — created in the Kendo bid builder |
| Expenses · Actuals · Invoices | ❌ UI-only — inline-grid / derived in the UI |
| Crew rates / Financials (Rate · P&W · OT · DT) | ❌ UI-only |
User creation (addUser) |
❌ admin-UI only (getUsers etc. are read-only) |
| Custom-Key definitions | ⚠️ values via updateJob&customKeys= only persist for pre-defined keys |
The API does cover the creative pipeline (jobs/shows/shots/assets/tasks/versions/renders/elements) plus timecards · schedule · contacts · resources · exchange-rates. The financial loop that is NIM's signature feature is the one part you can't automate — the design lesson behind expanding our clone to expose all of it as first-class tools.
The cross-tracker contract — project_summary
Identical shape to the other four servers (this is what the hub consumes):
{
"tracker": "nim",
"project": { "name": "NIGHTJAR", "id": 1, "code": "26000" },
"counts": { "sequences": 1, "assets": 1, "shots": 1, "tasks": 1 },
"sequences": ["SEQ010"],
"shots": { "SH010": { "sequence": "SEQ010", "tasks": { "ANIM": "review" } } },
"assets": { "HERO_CHAR": { "tasks": {} } }
}
NIM's per-entity statuses map to the canonical todo / wip / done / review / approved.
NIM quirks this server normalizes
- Inconsistent envelopes — bare array ·
{success,error,data}·{ID,success}(writes) ·[{success:false,error}]. - Generic list params —
getShots&ID=<showID>,getShows&ID=<jobID>,getAssets&ID=<jobID>; there is nogetTasks(a parent's tasks come fromgetTaskInfo&class=…&itemID=…). - Field-name gotchas — Shows carry
showname; task-status writes usetaskStatusID(camelCase). getUserJobslists only assigned jobs —list_jobsusesfindJobsfor the full list.
All verified live against NIM 7.2.9.
Part of the tracker-MCP quintet — hub · shotgrid · ftrack · kitsu · ayon. MIT · John Huikku.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。