plaky115-mcp
MCP server for the Plaky project management API. Provides 32 generated and 7 curated tools for spaces, boards, items, comments, and more, with safe defaults and dry-run mutations.
README
plaky115 · Python SDK + MCP server for Plaky
A clean, typed Python SDK and a Model Context Protocol server for the
Plaky public API — every documented operation, one
wheel, no surprises. Ported from the pinned TypeScript source
apet97/plaky115 (33ae2926, v1.0.11)
with behavior-level parity tests.
Plaky115 is unofficial and independent. It is not affiliated with, endorsed by, or sponsored by Plaky or CAKE.com. “Plaky” and “CAKE.com” are trademarks of their respective owners.
Install
pip install plaky115 # SDK only
pip install "plaky115[mcp]" # SDK + MCP server
Python 3.11+. One dependency stack: httpx2 + pydantic v2 (+ the
official mcp v2 SDK behind the extra).
Highlights
- All 32 public operations across nine resources — spaces, boards, items, comments, reactions, users, teams, item groups, item files — each pinned by contract descriptors and cross-surface parity tests.
- Async-first with a real sync client.
AsyncPlakyClientandPlakyClientshare pure logic only; the sync client never touches an event loop. - Safe under failure. Only GET requests retry (equal-jitter backoff,
bounded
Retry-After); writes make exactly one network attempt — even with an idempotency key. Mutation receipts recordattempted/mayHaveCommittedconservatively. - Strict wire contracts. Paged roots must be
{"data": [...], "hasMore": bool}; unsafe int64 JSON integers survive as exact decimal strings; bodies are bounded (16 MiB default / 64 MiB max); redirects are never followed;plk_keys are redacted everywhere. - Batteries for real workflows. Resolvers (exact-ID fast path, one
bounded list for text refs), field builders, bounded chunk readers with
exact
{page, index}continuation, deterministic spreadsheet-safe CSV/JSONL export, bulk updates with durable receipts. - Deterministic generation. OpenAPI contract → models and raw MCP
tools are committed, drift-checked artifacts;
scripts/generate.py --checkfails CI on a single changed byte.
SDK quick start
from plaky115 import AsyncPlakyClient
async with AsyncPlakyClient(api_key="...") as plaky:
page = await plaky.spaces.list(page_size=50)
for space in page.data:
print(space.id, space.title)
from plaky115 import PlakyClient
with PlakyClient(api_key="...") as plaky:
for space in plaky.spaces.iterate(page_size=100):
print(space.id, space.title)
The API key is sent only in the X-API-Key header, only to the configured
HTTPS origin. Python surfaces are snake_case; wire names stay Plaky
camelCase via model aliases.
MCP server
export PLAKY115_API_KEY=... # environment only; never an argument
plaky115-mcp --transport stdio # local hosts
plaky115-mcp --transport streamable-http --port 8000 # deployment
- Safe by default:
curatedmode +readscope. Write and destructive tools mount only with explicit flags; mutation workflows default to dry-run. - Two tool surfaces: 32 generated raw tools (one per operation) and 7
curated tools with 11 workflow IDs (
workspace.map,items.search,comments.thread,export.items,items.create,items.updateFields,comments.add,itemGroups.create,itemGroups.update,itemFiles.upload,itemFiles.update). - Structured everything: every tool publishes success and error output schemas; known failures return a typed envelope with attempt state; results are capped at 128 KiB with exact continuations.
- Hardened HTTP: stateless Streamable HTTP with request-scoped SSE,
36 MiB body cap, DNS-rebinding protection, loopback-default binding,
secret-free
/healthz. Uploads accept canonical base64 + metadata — never local paths.
Host config example: examples/mcp_stdio.json.
Deployment notes: docs/mcp.md and
docs/deploy-cloudflare.md.
Documentation
| Guide | Contents |
|---|---|
| docs/sdk.md | Clients, options, resources, errors, pagination |
| docs/mcp.md | Modes, scopes, result contracts, deployment |
| docs/compatibility-inventory.md | All 32 operations mapped to SDK + MCP names |
| docs/api-behavior.md | Wire-level behavior notes |
| docs/security.md · SECURITY.md | Credential and transport policy |
| docs/contract-evolution.md | fetch → diff → accept → regenerate |
| docs/live-certification.md | Read/write live gates |
Development
uv sync --all-extras --group dev
uv run pytest # 309 tests
uv run python scripts/verify.py --offline # full release-grade gate
The offline verifier checks: parity inventories, contract + generation
determinism, format/lint, strict pyright, tests with branch coverage,
example syntax, docs gates, wheel/sdist build, twine, fresh-environment
package smoke (base install without mcp, installed-wheel typing proof),
four-scope secret scan, and lock integrity. The read-only live
certification exercises all 17 read operations across four independent
surfaces.
License
MIT — see LICENSE. Retains the upstream copyright notice from the pinned source.
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。