The open marketplace for agentic workflows
Find, install, and publish workflows, harnesses, rules, and skills for Claude Code, Cursor, Codex, and any agent runtime.
npx openagents-cli add openagents/pr-reviewerWorkflow
A multi-step, goal-directed procedure an agent executes.
6 packagesHarness
Scaffolding around an agent: loop control, tool wiring, eval hooks, guards.
7 packagesRules
Constraints / style / policy files an agent must obey.
5 packagesSkill
A reusable capability with its own instructions + helper scripts.
Featured packages
Browse all →openagents/agent-guardrails
Agent Guardrails
Pre-action checks for destructive ops, budget/turn limits, confirmation gates, and a structured audit log
openagents/dependency-upgrader
Dependency Upgrader
Audits outdated dependencies, upgrades in risk order, runs tests after each, and opens one PR per major bump
openagents/pr-reviewer
Pull Request Reviewer
Structured PR review with risk classification, a correctness/security/test/perf checklist, and severity-tagged output
openagents/repo-onboarding
Repo Onboarding
Map an unfamiliar codebase into an architecture summary, entrypoints, conventions, and run/test instructions
openagents/secure-coding-rules
Secure Coding Rules
CLAUDE.md-style rules for secrets, input validation, authz, dependency hygiene, and PII logging
openagents/test-writer-loop
Test Writer Loop
Red/green loop that picks an untested module, writes a failing test, implements it, and iterates with coverage deltas
Most installed this week
See all →openagents/hello-world
Hello World
A minimal workflow that greets you, inspects the repo, and summarizes it.
openagents/pr-reviewer
Pull Request Reviewer
Structured PR review with risk classification, a correctness/security/test/perf checklist, and severity-tagged output
openagents/eval-harness
Agent Eval Harness
Build a real eval suite for an agent or prompt: cases, graders, a baseline, and a regression gate
openagents/typescript-style-rules
TypeScript Style Rules
Rules an agent must follow when writing TypeScript: type honesty, error handling, module boundaries, and what not to abstract
Browse by tag
All tags →How it works
Find
Search or browse workflows, harnesses, rules, and skills built by the community.
Install
Run one command. OpenAgents drops the package into the right place for your runtime.
Run in your agent
Claude Code, Cursor, Codex, and other agent runtimes pick it up automatically.
Publish yours
Free packages are always free to publish and free to install — no platform fee, ever. Charge for a package and keep 90% of every sale; OpenAgents takes a 10% platform fee to cover payments and hosting.