OSOP now supports 18 AI coding platforms. This post explains what OSOP does, how it integrates with each platform, and why a universal workflow protocol matters.
What is OSOP?
OSOP (Open Standard Operating Procedures) is two YAML formats:
.osop— .osop — a workflow definition. Describes what should happen: nodes (16 types), edges (13 modes), security metadata, approval gates..osoplog— .osoplog — an execution record. Describes what actually happened: timestamps, tool calls, token counts, AI reasoning, human decisions.
Together they form a closed loop: define, execute, record, optimize.
Why Does This Matter?
Without OSOP, every AI coding session is a black box. The agent did "something" and you got a result. If it went wrong, you dig through chat logs. If it went right, you cannot reproduce it. There is no audit trail, no optimization data, no way to share the workflow with a colleague using a different tool.
OSOP changes this. Every session produces a structured, portable record that works across all 18 platforms.
Platform-by-Platform Integration
Tier 1: Native Integrations
These platforms have first-class OSOP support with skill systems or plugins.
| Platform | Integration | How to Install |
|---|---|---|
| Claude Code | Native plugin with 5 slash commands | claude /install-plugin https://github.com/Archie0125/osop-skill |
| OpenClaw | 5 skills on ClawHub registry | clawhub install osop |
| GitHub Copilot | MCP server integration | pip install osop-mcp |
Tier 2: Rule File Integrations
These platforms read markdown instruction files. Drop in one file and OSOP works.
| Platform | File | Location |
|---|---|---|
| Cursor | .mdc rule | .cursor/rules/osop-session-logging.mdc |
| Codex | AGENTS.md | Project root |
| Windsurf | .md rule | .windsurf/rules/osop-session-logging.md |
| Continue.dev | YAML rule | .continue/rules/osop-session-logging.yaml |
| Aider | CONVENTIONS.md | Project root |
| Cline | .md rule | .clinerules/osop-session-logging.md |
| Roo Code | .md rule | .roo/rules/osop-session-logging.md |
| Devin | Playbook | playbooks/osop-session-logging.md |
| Zed | .rules | Project root |
| Sourcegraph Amp | AGENT.md | Project root |
| Trae | project_rules.md | Project root |
Tier 3: Config-Based Integrations
| Platform | Integration |
|---|---|
| Obsidian | Custom prompt in .obsidian/copilot-custom-prompts/ |
| PearAI | Custom slash command in config.json |
| Sweep AI | SKILL.md in .claude/skills/ |
| SWE-agent | YAML config with system_template |
Universal Installer
Do not want to figure out which file goes where? The universal installer handles it:
git clone https://github.com/Archie0125/osop-agent-rules.git cd osop-agent-rules ./install.sh # auto-detects your tools ./install.sh --all # installs for everything
The installer checks for config directories (.cursor/, .windsurf/, .clinerules/, etc.) and only installs for tools you actually have. It never overwrites existing files unless you pass --force.
What You Get
After your AI agent completes a multi-step task, it automatically produces two files:
sessions/2026-04-01-fix-auth-bug.osop # What should have happened sessions/2026-04-01-fix-auth-bug.osoplog.yaml # What actually happened
You can:
- Visualize — drag both files into the editor for an interactive node graph
- Analyze — run risk analysis to find missing approval gates or destructive commands
- Optimize — compare execution logs across runs to find slow steps and failure hotspots
- Audit — structured evidence of every AI action, tool call, and human decision
- Share — portable YAML files that work across any OSOP-compatible tool
The OSOP Ecosystem
| Component | What It Does |
|---|---|
| Spec v1.0 | JSON Schema for .osop and .osoplog formats |
| Visual Editor | Interactive node graph, risk analysis, execution replay |
| Python CLI | pip install osop — validate, render, run |
| JS/TS SDK | Parse and manipulate .osop files in Node.js |
| Python SDK | Python library for workflow creation and validation |
| 39+ Examples | DevOps, AI agents, data pipelines, incident response |
| Agent Rules | Drop-in files for 18 platforms |
| ClawHub Skills | 5 skills for OpenClaw ecosystem |
| MCP Server | Model Context Protocol for Copilot, Cursor, Claude |
Get Started
# Install the CLI pip install osop # Or use the agent rules for your AI tool git clone https://github.com/Archie0125/osop-agent-rules.git cd osop-agent-rules && ./install.sh # Or install as a Claude Code plugin claude /install-plugin https://github.com/Archie0125/osop-skill # Or install on OpenClaw clawhub install osop
OSOP is open source (Apache 2.0). The spec, SDKs, editor, and all integrations are free. We want OSOP to be as boring and ubiquitous as JSON Schema.