Infrastructure for the agents that write your code.
Rness keeps your organization's standards, decisions, specs and plans in one place and writes them into every repository's AGENTS.md. Claude Code, Codex, Cursor and GitHub Copilot already read it.
npm create rnessOpen source, MIT · Runs on your machine · No new agent, workflow or configuration format
AI development is distributed. Organizational knowledge shouldn't be.
Every agent and every repository holds a different piece of what your organization knows, copied by hand and drifting apart.
- 01
Fragmented context
Different agents and repositories contain different pieces of organizational knowledge.
- 02
Cross-project drift
Rules and standards get copied between repositories and become inconsistent.
- 03
Lost decisions
Architecture decisions, plans and specifications drift away from the agents implementing them.
- 04
No organizational visibility
Engineering leaders cannot see what governance applies to which repository.
3 repositories4 instruction formats0 shared source of truth
A governance layer above your agents.
Rness does not replace your agents or redefine their configuration. It provides the organizational layer that connects them.
- Claude, by Anthropic
- GPT, by OpenAI
- Gemini, by Google
- …
- Claude Code
- Codex
- Cursor
- GitHub Copilot
- …
- GitHub
- GitLab
- Bitbucket
- Jira
- CI/CD · …
Your Git provider
Remains the source of code. Rness sits beside it and changes nothing about how you host, review or ship.
- code
- history
- pull requests
.rness
Becomes the source of organizational truth: one repository in your organization, versioned like code.
- standards/rules, by scope
- adr/architecture decisions
- specs/ · plans/what is built, and how
- skills/reusable agent skills
Agents execute. Rness governs.
One organization. Every repository. Every agent.
Define shared context and governance once. Rness applies it across every repository and every agent your teams use.
One organization. Every repository.
Define shared context and governance once and apply it across every repository. The repositories stay independent: Rness provides the organizational layer above them.
- GitHub
- GitLab
- Bitbucket
Define once. Govern every agent.
Claude Code remains Claude Code. Codex remains Codex. Cursor remains Cursor. Rness writes the shared layer into the files they already read.
- AGENTS.md
- CLAUDE.md
- No new format
The right context, at the right scope.
Organization-level context is inherited by every repository. Each repository, and each scope inside it, adds only what is specific to it.
- Organization
- Repository
- Directory
Your agents understand why, not only what.
Architecture decisions, specifications and plans are linked, versioned beside the code and moved on your board as the work happens.
- ADR
- Specs
- Plans
- Boards
Turn engineering standards into organizational policy.
Policies carry a scope: organization, repository, directory or file pattern. Native agent configuration files stay where they are.
- Organization
- Repository
- Directory
- File pattern
Product in action
Your agent works. Your board keeps up.
A developer writes a specification with Claude Code, approves it and ships it. Agent Pulse, the board Rness keeps in step with .rness on GitHub Projects, GitLab or Jira, moves the card as it happens.
- In app1, a developer starts Claude Code; rness reports the scope and its standards and decisions.
- They run /rness:spec. Claude reads the security standard and an ADR, writes specification 0012 as Draft, and its card appears on Agent Pulse, marked working.
- They run /rness:plan from 0012: that approves it, the card moves to Approved, and Claude writes the plan and builds it; tests pass. /rness:plan check 0034 checks every task and the card moves to Implemented.
- The session ends: the working mark is cleared and the session stays in the card's history.
Claude remains Claude. The developer stays in the same workflow. Rness provides the organizational governance around it.
Nothing changes for developers.
Rness works around your workflow, not against it. Three steps, then everyone keeps coding.
- 01
Connect
Connect your GitHub, GitLab or Bitbucket organization. On Atlassian, the board lives in Jira.
- 02
Govern
Define organizational context, standards, policies, decisions, specs and plans.
- 03
Keep coding
Developers keep using Claude Code, Codex, Cursor, GitHub Copilot or other agents exactly as before.
Governance you can run from the terminal, and in CI.
- rness create
- Create or join an organization, or start blank
- rness add
- Add a repository to the workspace
- rness sync
- Sync organizational governance
- rness context
- The resolved context of a scope
- rness status
- Where every decision, spec and plan stands
- rness board push
- Put the work on your board
- rness validate
- Check the context and its blocks
- rness login
- Reach private repositories
rness sync --check writes nothing and exits 1 when a block is out of date: drift is caught in CI, before it reaches an agent.
A new layer in the AI development stack.
Coding agents are not what Rness competes with: they are what it governs. The relevant comparison is between the layers that address organization-wide AI governance.
Layer / capability | Coding agents Claude Code · Codex · Cursor · Copilot | Project rules CLAUDE.md · AGENTS.md · .cursor/rules | Agent governance Rness |
|---|---|---|---|
| Execute coding tasks | ✓ | - | - |
| Repository instructions | ✓ | ✓ | ✓ |
| Cross-repository governance | Limited | Limited | Core |
| Cross-agent governance | Limited | Limited | Core |
| Organization-wide context | Limited | Limited | Core |
| Policy enforcement | Limited | Partial | Corenot shipped yet |
| ADR / architecture memory | Partial | Partial | Core |
| Context inheritance | Partial | Partial | Core |
| Configuration drift | Limited | Partial | Core |
Categories, not vendors. Coding agents execute tasks. Project rules instruct a single repository. Agent governance spans the organization - every repository, every agent.
Give every AI agent the same understanding of your organization.
npm create rness