Open Canvas
A Workspace Built for Coding Agent Workflows
Open Canvas is an open-source, browser-based development environment that centralizes your data, dynamically allocates agent skills, and standardizes application execution across any tech stack. Build faster by eliminating the friction between you and your agents.
The Problem
Coding agents build applications in minutes. But developers spend more time preparing data, re-explaining context, and managing terminal sessions than the agent spends building.
Every new session starts from zero — no memory of previous work, no knowledge of where data lives, no awareness of project conventions. The agent reinvents the wheel every time.
Open Canvas solves this by making data accessible, skills persistent, and execution standardized — so every session builds on the last.
Capabilities
Built for the Agent-First Developer
Every feature in Open Canvas exists to reduce the friction between your data and your agents.
Centralized Shared Data
Link global data across projects without duplication. Every agent session accesses the same source of truth — no more copying files or hunting for the right version.
Dynamic Skills Allocation
Projects ship with default skills that agents refine over time. A built-in Dynamic Skills instruction tells the agent to create new skills as it identifies recurring patterns.
Standardized Run Scripts
Every app includes a run.sh file that handles the full stack — from dependency installation to multi-server startup. Press Run App once, regardless of the technology.
Parallel Agent Terminals
Run multiple agent sessions on the same project simultaneously. One agent building features, another running tests — Claude, Codex, or Gemini side by side.
Built-in Git Manager
Branch switching, commit history, remote management, and stash log directly in the workspace. Set git context without instructing an agent to do it.
Self-Optimizing Projects
Each project maintains its own CLAUDE.md and skill hierarchy that evolve as the agent works. The more sessions you run, the smarter the project becomes.
Workflow
Three Steps to Agent-Ready Projects
Create a Project
Start a new project or import an existing repo. Open Canvas scaffolds the skill hierarchy, run configuration, and data links automatically.
Link Your Data
Connect shared data folders — client files, reference docs, templates. Data is linked, not copied, so every project accesses the same source of truth.
Build and Iterate
Open agent terminals, run your app, and monitor progress — all in a single browser tab. Agents build skills as they work, compounding productivity over time.
Architecture
Data Strategy by Design
Open Canvas structures projects so agents always know where data lives and how to operate.
Skill Hierarchy
At the root level, Open Canvas maintains a global skill that evolves with the workspace. Each project has its own .claude/skills/ directory with project-specific automation — including a Dynamic Skills instruction that tells the agent to optimize itself over time.
Shared Data Layer
Global data is linked — not copied — into projects. Client briefs, reference documents, and templates exist in one place. When the source updates, every project reflects the change.
Run Configuration
Every project includes a run.sh script and a run-app.md instruction file. The agent reads the instruction, executes the script, or writes one if it doesn't exist — handling everything from simple HTML to multi-stack deployments.
Ready to Eliminate Data Friction?
Open Canvas is open source. Explore the repo, try the workspace, or read the deep dive on data strategy for coding agents.