mcpbeat Sign in

Canvas MCP Server

by avatargaia Your server? Claim it
answering

Canvas MCP is answering right now. Last checked 6 min ago. 92 installs a week from npm. It exposes 18 tools. Last commit 7 Oct 2026.

TeamAgent Canvas MCP(服务端 0.4.0):91 个工具——开岗、改配置、跑动作、看数据、发布并配知识库。可匿名试用,带 Key 解锁全部。

Installs per day peak 187 · avg 14 · +229% w/w
a month agotoday
Uptime history 51 days of history · worst day 95%
51 days agonow
100.0%
Uptime 24h
76 of 76 checks
18
Tools
read from the server
1446 ms
Response time
average over 24h
92
Installs / week
npm and PyPI

What changed 20

Every tool that appeared, vanished or quietly changed what it asks for. Recorded since 28 September 2026. No other catalogue keeps this.

20 Oct 4 tool descriptions were rewritten5 times that day list_kbs, publish_kb, read_kb and 1 more
20 Oct 2 tools changed the parameters they ask for publish_kb, read_kb
3 Oct 5 tools appeared hire_staff, list_kbs, publish_kb and 2 more
3 Oct 4 tool descriptions were rewritten hire_staff, patch_staff_config, publish_kb and 1 more
3 Oct 2 tools changed the parameters they ask for publish_kb, read_kb
3 Oct a tool changed version
28 Sep a tool changed version
and 1 more, back to 28 September 2026

Canvas MCP does not always answer

Over the last week it answered 99.0% of our checks. We check every 15 minutes, so you hear about the next outage within the hour — not from your users.

Three servers free · no card

Connect this server

Endpoint below is the one we actually reach during checks — not the one copied from a README. Last verified 6 min ago.

run in your terminal
claude mcp add canvas-mcp --transport http https://agent.avatargaia.top/api/mcp/canvas
Settings → Connectors → Add custom connector, then paste this address
https://agent.avatargaia.top/api/mcp/canvas
~/.codex/config.toml
[mcp_servers.canvas-mcp]
url = "https://agent.avatargaia.top/api/mcp/canvas"
.cursor/mcp.json
{
  "mcpServers": {
    "canvas-mcp": {
      "url": "https://agent.avatargaia.top/api/mcp/canvas"
    }
  }
}
.vscode/mcp.json
{
  "servers": {
    "canvas-mcp": {
      "type": "http",
      "url": "https://agent.avatargaia.top/api/mcp/canvas"
    }
  }
}

Available tools 18

Read directly from the server with tools/list, grouped by what they act on. If a tool disappears, we record the date.

staff
get_staff_config
读岗位配置:personaPrompt / serviceFlow / promotionConfig / productSources / contextMode / layoutStyle / templateId / llmMode(只说是否 BYOK,不给密钥)/ tts。参数:slug 必填。改之前先读,否则会用"想当然的人设"覆盖客户调过的版本。
list_my_staff
列出我名下的数字员工岗位(slug / 名称 / 模板 / 画布链接)。用户问"我有哪些岗位"用它;后续所有按岗位操作的工具都要 slug,slug 一律从这里拿,不要凭记忆拼。参数:无。
set_staff_kbs
设定一个岗位用哪些知识库(整体替换)。参数:slug 必填;kbs 是库名数组:我名下的库写库名;工作区的库或别人共享的库写 list_kbs 返回里的 ref(如 ws:xxx/库名)。别人的库要授权时勾了「可对外」才配得上,不满足的会被忽略并在 kbDropped 里说明;传 [] = 这个岗位不用任何库。改完立刻生效。
actions
list_actions
列出这个岗位(按其模板)能做的动作:id / title / description / cost / credits / anonymous / job / params(JSON Schema)。参数:slug 必填。调 canvas_action 前先调它,不要猜 action 名——各岗位动作集不同,params 按它给的 schema 填。
canvas
canvas_action
执行一个岗位动作(如 AICare 的 preflight / list_checks / gen_report / run_detection / resolve_user)。参数:slug、action 必填;params 按 list_actions 的 schema 填;session_id 想让结果落在某次会话里就带;hostUserId 宿主用户标识可选。返回 {ok, mode:"sync", data} 直接用;{ok, mode:"job", jobId, pollSec} 用 get_job 轮询。代价:list_actions 标 paid 的会从岗位出资人余额扣 credits(出报告 5 / 跑检测 2),失败不扣;同键(如同对象同一天)重复调用返回缓存结果(cached 标记)不重复扣,用户催第二次可放心重试。典型编排:list_actions → canvas_action(preflight) 拿背景 → canvas_action(gen_report) → get_job → 用自然语言讲给用户。公共岗位:gaia-academy(龙虾学院 · Agent 进修)的动作标 public:true——任何持有效 Key 的调用方都能跑,不需要拥有该岗位;学员身份取调用方 Key 的用户,非 public 动作对非归属方仍 403。典型用法:canvas_action(gaia-academy, search_courses → start_exam → submit_exam → get_principle)。
events
query_events
跨会话按类型拉事件(倒序):type = chat_user / chat_assistant / track / action_call / page_open / push。参数:slug 必填;type、from/to、limit(≤500)。每条带 source(mcp / page / server,看得出谁发起的);action_call 的 payload 有 params/ok/ms/credits/cached/jobId——计费与排障的证据链;push 的 payload 有 preview。需要有效 Key,匿名不开。
hire
hire_staff
开一个新的数字员工岗位。参数:templateId 必填(list_templates 里 canOpen=true 的,如 kb-assistant 知识库问答员);name 员工名字;kbs 开岗时就配上的知识库。返回 slug、画布链接 canvasUrl(发给顾客 / 挂服务号的就是它)和实际配上的知识库 kbs(名下没有的库名会被忽略,以返回为准)。代价:新岗位先免费试用几天,之后按月从主人余额扣 Token,对话按次另计——开之前跟用户确认;每个账号岗位数有上限,到上限会报错,让用户先删掉不用的。
job
get_job
按 job_id 查长任务:queued / running / done(带 result) / failed(带 error)。建议每 3 秒查一次。error 是人话(如"关爱对象不存在或已删除"),照原话转述给用户,不要改写成"系统繁忙",也不要拿同参数重试。
kbs
list_kbs
列出我名下已发布的知识库(库名 / 版本 / 发布时间 / 篇数 / 每页的 path 与标题),以及我名下每个岗位各自配了哪些库;还有 workspaces(我所在工作区自己的库 + 工作区引用的库)和 shared(别人授权给我的库),这些库各带 ownerKey 与 ref。在某个工作区里干活时,优先读那个工作区的库和它引用的库,说法冲突时都列出来。用户问"我有哪些知识库""这个岗位用的什么资料"时用。知识库是数字员工回答业务问题的依据:岗位配了库,访客问到时它会先查库再答。
open
open_canvas
为岗位生成画布链接 + 会话 ID。用于用户需要"看得见的界面"(看报告/看商品/做检测)。参数:slug 必填;hostUserId 宿主用户标识(强烈建议带,会话归属和离线补推都靠它);targetId 关爱/业务对象;app 指定画布页(如 aicare-kf)。返回 {url, session_id}——把 session_id 存下来,get_session_log / push_screen 都用它。会话在返回时已建立:现在就能 push_screen 给"还没来的人"备卡片,用户打开链接即补推;人打开后聊天/埋点都归到同一个 session_id。
patch
patch_staff_config
局部修改岗位配置,patch 只放要改的字段:personaPrompt(人设) / serviceFlow(服务流程,一行一步) / promotionConfig(只放要改的键,其余保留:categories:[{name,priority,url}] 主推;look:{brandName,slogan,intro,avatarUrl,theme(ember|green|blue|purple),particles,startText,footer,placeholder,emptyTitle,emptyHint,showPhoto} 画布外观与文案;sell:{sop,pdd} 是否推荐服务包 / 拼多多商品;kbs:[库名] 用哪些知识库,也可用 set_staff_kbs)/ productSources([{key,label,searchUrl,weAppId?,weAppUsername?}]) / layoutStyle(split|pip|classic) / contextMode(inline|webhook) / contextWebhookUrl / name。参数:slug、patch 都必填。这是写操作,改的是客户的岗位,改完立刻生效;不确定用户真要改时先复述一遍再调。
publish
publish_kb
把一套整理好的 Markdown 发布成知识库(发布到我名下;带 slug 则同时配给那个岗位;带 owner=ws:<工作区id> 则发布成那个工作区的库,要求我是该工作区 owner/admin)。结构:wiki/index.md 是总目录(每篇一行:链接 + 一句摘要 + 更新日期,数字员工靠它挑页,摘要要准);其余 wiki/<主题>/<文章>.md,一个概念一篇,标题用访客会问的说法,单篇别超过 6000 字。页面要带图片 / 视频 / 三维模型 / 外链时,在页里写一个 ```kb-media 围栏,一行一个「类型 | 标题 | 链接 | 说明」,类型 image / video / model(.glb) / link / embed(只认白名单站点),链接用 https:// 或本站 /static/ /uploads/ /apps/ 路径——素材只放链接,不要把文件内容塞进来;数字员工回答时会把素材显示给访客。库主人可以开「只用已校验的页」,那时你新发或改过的页都是待校验,要人在网页上核对后才对访客生效,你不能自己批。规矩:①数字、日期、引文必须照原始资料逐字写,不改写不推算;②只收 wiki/ 下的 .md,原始资料不要发上来;③同名再发是整库替换,更新前先 read_kb 把现有页面读回来合并;④发布等于公开——岗位是匿名访客都能聊的,个人隐私、客户资料、令牌、内部运维内容一律不放(平台会拦身份证号和令牌,其它靠你自检)。checked 必须为 true 才会发。上限 200 个文件 / 单个 64KB / 合计 2MB。参数示例:{"name":"选课指南","checked":true,"files":[{"path":"wiki/index.md","content":"# Knowledge Base Index\n..."},{"path":"wiki/xuanke/rules.md","content":"# 选课时间与学分上限\n..."}]}(files 是数组,每项一个 {path, content},不是 {路径: 内容} 的对象)。这是写操作,发之前跟用户确认库名和要不要配给岗位。
push
push_screen
把内容推到某个画布会话的屏上。参数:session_id 必填(来自 open_canvas 或 list_sessions);content 必填,写成一条"助手回复"文本,可含 ```canvas-md(Markdown 卡片)/ ```canvas-html(完整 HTML 文档)/ ```canvas(JSON 块)围栏,围栏外文字显示为字幕;speak=true 朗读字幕;title 可选。会话在线立即上屏;不在线(含还没人打开过的)存为待展示,该会话或同岗位同一 hostUserId 下次打开自动补推。屏属于会话不属于人:别复用别人的 session_id;要执行有副作用的动作(出报告/跑检测)用 canvas_action。需要有效 Key,匿名不开。
read
read_kb
读一个知识库。参数名是 name(库名,就是 list_kbs 返回里的 kbName)和 page(页面路径,取自页面清单里的 path)。只传 name → 这个库的版本和页面清单;再传 page(如 wiki/index.md)→ 那一页的原文。要更新一个库之前先用它把现有页面读回来——publish_kb 是整库替换,没带上的页面会消失。读工作区的库或别人共享的库要加 owner(list_kbs 返回里的 ownerKey,如 ws:xxx)。
session
get_session_log
单个会话按时间正序的完整流水:用户消息 / AI 回复(带 tokens) / 埋点 / 动作调用 / 推屏。参数:slug 和 session_id 两个都必填(slug 用于显式鉴权)。调用前先从 list_sessions 或 open_canvas 拿到合法 session_id。隐私边界:只能读自己岗位的记录。需要有效 Key,匿名不开。
sessions
list_sessions
岗位的访客会话列表(最近优先):session_id、宿主用户标识、轮数、时间。参数:slug 必填;q 按宿主用户标识模糊搜;from/to;limit;cursor 翻页传上一页返回的 nextCursor。想看"最近都服务了哪些人"用它,再拿 session_id 看明细。需要有效 Key,匿名不开。
stats
get_stats
岗位一段时间的汇总:会话数、真人对话轮数、埋点数、动作调用数、tokens、按日曲线。参数:slug 必填;from/to ISO 时间,默认最近 30 天。用户问"这岗位最近怎么样/今天多少人来"就用它,不要靠聊天历史猜数字。需要有效 Key,匿名不开。
templates
list_templates
列出可开岗的数字员工模板。用于:想知道"系统能开哪些岗位"、或用户要开通新岗位时。返回每个模板的 templateId / 名称 / 说明 / canOpen / missingHandlers / 动作清单。别用它找"已开好的岗位"——那是 list_my_staff。canOpen=false 的模板不可开通(缺实现),别推荐用户选。参数:无。

Endpoints

URLTransportStateLatencyChecked
https://agent.avatargaia.top/api/mcp/canvas streamable-http answering 1060 ms 6 min ago

Alternatives to Canvas MCP

same job, measured the same way
B
Brilliant
by brilliant

Local MCP bridge: your coding agent designs on Brilliant's real vector canvas, fully on 127.0.0.1.

117 installs/wk local only
Infiknit
by synthenova

Local Infiknit MCP for image and video canvas workflows. Keep Infiknit open on this computer.

local only
Modellix Agent Canvas
by modellix

Local MCP canvas for Modellix image generation, editing, task recovery, and visual project work.

77 installs/wk local only
Flato Design MCP
by sologovision

Connect AI agents to Flato's editable canvas runtime through a hosted MCP server.

answering
Klyaksa
by klyaksa

Online whiteboard your AI agent draws on: live shared canvases, solo or with your team.

answering
Memory
by humaux

Persistent long-term memory for AI agents: semantic search, knowledge graph, and task canvas.

answering
Canvas Lms MCP
by bruchris

TypeScript MCP server for Canvas LMS — 187 tools across 42 domains.

2 573 installs/wk local only
Canvashand
by canvashand

Figma MCP server and plugin: AI agents read and edit live Figma files, no Figma API token.

local only

Canvas MCP — questions

Answers built from our own checks of this server.

What can Canvas MCP do?
It exposes 18 tools, read directly from the server on our last check. Among them: canvas_action, get_job, get_session_log, get_staff_config, get_stats, hire_staff and 12 more. The full list with descriptions is on this page — we take it from the server itself via tools/list, not from a README. How MCP servers expose tools in the first place →
Is Canvas MCP working right now?
We send a real MCP handshake every 15 minutes. Over the last 24 hours 76 of 76 checks got a reply (100.0%), average response time 1446 ms. The bar chart above shows every period we have measured.
How do I connect Canvas MCP?
Copy the ready config from this page. We generate it for Claude Code, Claude Desktop, Codex, Cursor and VS Code, each in the form that client actually reads. It is a remote server, so there is nothing to install — the client connects to the address.
Does Canvas MCP need an API key?
No. Canvas MCP completed a full MCP handshake with us as an anonymous client and listed its tools without asking for anything. All 18 of them are readable on this page. This is what we observed, not what the docs claim.
How fast is Canvas MCP?
It answers our handshake in 1446 ms on average, which is faster than 1% of all working MCP servers we measure. That is on the slow side — worth knowing if the tool sits inside an interactive loop. The comparison comes from our own checks across the whole registry, every 15 minutes.
How many people use Canvas MCP?
The npm package @avatargaia/canvas-mcp was installed 92 times in the last week. Week over week that is +229%. We show installs rather than GitHub stars on purpose: a star is a bookmark, an install is someone actually running it.
Is Canvas MCP open source?
The code is open source: it is published under the MIT licence, written in JavaScript and 0 stars on GitHub. The licence covers the code in the repository, not the hosted service at agent.avatargaia.top, which runs on its operator's own terms.