vibe-canvas
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@vibe-canvasopen a thinking canvas"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Vibe Canvas
Simplified Chinese | English
Let Codex conversations grow into a filterable thinking canvas on the right side in real time.

Introduction
Vibe Canvas is a local-first Codex plugin. You continue chatting with Codex normally on the left, and the Browser on the right automatically displays three layers of content:
The verbatim conversation transcript
AI-summarized structured understanding
A continuously growing mind map as the conversation progresses
It is not a standalone website that requires you to input again, send again, or manually save. The HTML page is just a real-time projection view on the right side of Codex; input, understanding, and reasoning still happen within the current Codex task.
Ideal for brainstorming, solution exploration, study notes, retrospectives, requirement analysis, and any thinking process that benefits from "seeing structure while chatting."
Related MCP server: repo-memory-mcp
Features
Conversation as input: The left Codex is the only text entry point; there is no second input box inside the canvas.
Original text traceable: Every user input is preserved verbatim; AI rewrites are not presented as the original.
AI auto-pick: New content is classified as
picked,candidate, orignored.Candidate buffer: Content that cannot be determined immediately waits for subsequent context, not rushed into the structure.
Human final decision: Each turn has a Pick Checkbox; human selection always overrides later AI judgments.
Real-time restructuring: When a Pick is removed, the corresponding content immediately exits the structure and mind map without calling the model again.
Local-first: Session and theme preferences are saved in the
.local/directory of the current workspace.Dual themes: The bright blue default theme and a warm theme can be toggled instantly with preference remembered.
Incremental sync: Each turn only reads the compact projection and writes the current increment, avoiding repeated replay of the entire conversation.
Screenshots
Pick Selected and Unselected
Left side shows green selected state, right side shows gray unselected state; removing the pick updates the structure and mind map synchronously.

Requirements
Codex Desktop, with support for local Plugin, Skill, MCP Server, and right-side Browser
Node.js 20 or higher
A working
codexCLI
Current version developed and verified on macOS.
Installation
Register the GitHub repository directly as a Codex Marketplace, then install the plugin from it:
codex plugin marketplace add prozacLaputa/vibe-canvas
codex plugin add vibe-canvas@vibe-canvas --jsonConfirm installation status:
codex plugin list --jsonAfter installation or upgrade, it is recommended to create a new Codex task to let the Skill and MCP tools rediscover.
Usage
In a new Codex task, enter:
使用 $vibe-canvas 打开一个思考画板Then chat normally:
Codex opens Vibe Canvas on the right side.
Before each response, the current conversation syncs the original text and structural increment of that turn to the canvas.
The AI automatically selects explicitly valuable content, placing uncertain content into the candidate buffer.
You can click Pick in the original text area to decide at any time whether a turn enters the structure and mind map.
The Switch in the top right corner toggles between blue and warm themes.
The canvas activation scope is limited to the current Codex task; ordinary tasks are not automatically taken over.
How It Works
flowchart LR
U[左侧正常对话] --> C[当前 Codex 任务]
C -->|读取紧凑投影| M[本地 MCP Server]
C -->|写入本轮原文与结构增量| M
M --> J[本地 Session JSON]
J -->|revision 变化| B[右侧 Browser 投影]
B -->|Pick / 主题 PATCH| JThe core consists of three parts:
Skill handles timing and constraints: After explicit activation, each turn reads the compact context via
vibe_canvas_get_projectionand writes the current increment viavibe_canvas_sync_turn. It does not start a second Codex session or fake user messages.MCP Server handles state: A Node.js process provides both MCP tools and a local HTTP API listening only on
127.0.0.1, persisting the session as JSON.Browser handles projection and filtering: The page checks the internal revision approximately every 900 ms and only re-renders when the state changes. Pick and theme switching go directly through local HTTP, no model calls, no extra tokens consumed.
Tokens mainly come from one compact structure read and one small incremental write per turn; browser polling, theme switching, and manual picks are all local operations.
Default data location:
<workspace>/.local/vibe-canvas-demo/sessions/<session-id>.json
<workspace>/.local/vibe-canvas-demo/preferences.jsonShortcuts
No additional shortcuts currently; use the Codex conversation and the Pick / theme controls in the canvas.
Development
npm testTests cover opening the canvas, incremental sync, AI pick, candidate re-evaluation, human override, theme preference, and state restoration via public MCP JSON-RPC and local HTTP projection API.
Changelog
See CHANGELOG.md.
License
MIT © 2026 prozacLaputa
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Servers
- AlicenseAqualityCmaintenanceLocal-first external brain for Claude Code, Codex, and any MCP client. Stores decisions, entities, and session artifacts in one SQLite file and exposes MCP tools for recall, page, promote, review, graph-query, and source-status.11154MIT
- AlicenseCqualityCmaintenanceMCP server for sharing source-backed engineering memory across AI coding clients like Cursor and VS Code.301MIT
- Alicense-qualityCmaintenanceA local MCP server that provides a shared context and learning foundation across multiple AI tools (Claude, Copilot, Codex) for multiple projects, enabling persistent knowledge, decisions, and gap reflection through note storage.MIT
- AlicenseBqualityCmaintenanceLocal Markdown-backed memory tools for Codex and other MCP-capable agents. Exposes durable agent knowledge via CLI and MCP server.5MIT
Related MCP Connectors
AI Reasoning Cache & Consensus Layer with 11 MCP tools via Streamable HTTP.
Private-by-default, local-first memory/context/task orchestrator for MCP apps and agents.
Hosted MCP memory: save sessions/decisions once, search from Claude, Cursor, ChatGPT. EU-hosted FTS.
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/prozacLaputa/vibe-canvas'
If you have feedback or need assistance with the MCP directory API, please join our Discord server