One pipeline. Every agent.
Stop correcting your AI. Start directing it. reveren is structured, versioned, repo-aware guardrails for AI coding agents — protocols plus pipelines that make whichever agent you already pay for dramatically better. Author once. Sell to every agent.
npx @reveren-ai/core init✓ Detected: Next.js 16 · TypeScript · pnpm · Vitest✓ Selected agent: Claude (via Cursor)↳ Writing .protocols/ · plan-product.protocol.md · plan-engineering.protocol.md · ship.protocol.md · review.protocol.md↳ Writing .reveren/config.ts↳ Wiring package.json scripts✓ Ready. 12 protocols installed.reveren run plan-engineering→ The agent now reads your repo before it touches it.
What ships, when.
The CLI and protocols are live today. The Pod Marketplace and hosted runtimes are next.
Protocols + pipelines
Author versioned, repo-aware protocols. Chain them into deterministic pipelines. Run from any agent.
- Plain-Markdown .protocols/ format
- Multi-step pipelines
- MIT-licensed CLI
- Bring your own model
Pods + Marketplace
Bundle protocols into agent-bound pods. List on the Pods Marketplace. 70/30 creator split.
- Author once
- Sell to every agent
- $1 floor · 70/30 split
Hosted pods + knowledge graphs
24/7 always-on agent runtimes. Knowledge graphs and pruning as opt-in installables.
- Always-on pods
- Knowledge graph add-ons
- MCP connectors
- EA / PA pod templates
Three audiences, one operating manual.
reveren works the same way whether you're shipping production software in a team of fifty, building your fourth side project this quarter, or governing five engineering pods running three different agents.
Your AI agent writes code that drifts from your conventions, your testing patterns, your migration strategy. You spend the second hour of the day correcting it. reveren makes the agent inherit the team's protocol on day one.
You're shipping with Lovable, Bolt, v0, or Cursor. The agent moves fast, but the third feature breaks the first two. reveren turns your scattered prompts into versioned guardrails that keep the agent on rails as your project grows.
Your team pays for three different agents. Quality and review burden vary by who's on duty. reveren is the standards layer — same protocols, same pipeline, every developer, every agent. One thing to govern instead of five.
Four primitives. Composable in any order.
Protocols describe the work. Pipelines chain them. Standards make them consistent. Audit makes them defensible. Together they're the operating manual the agent didn't ship with.
Protocols
Versioned, repo-aware instructions in plain Markdown. Author once. Run from any agent. The format is open — your work travels with you.
Pipelines
Chain protocols into deterministic workflows: plan → ship → review → qa → document. The agent stops freelancing and follows the manual.
Standards
One source of truth for naming, testing, coverage thresholds, deploying, reviewing — applied identically by every agent on every PR. Coverage drift becomes a CI failure, not an oversight.
Audit
Every run is recorded. Every decision is traceable. When the regulator, the post-mortem, or the new hire asks why, the answer is in the log.
Works with whichever agent you already pay for.
Bring your own model. reveren is the standards layer between your codebase and whichever coding agent you use today. Switch agents next quarter and your protocols come with you.
Claude
Cursor
Copilot
Windsurf
Lovable
v0
mrktable runs on its own protocols.
mrktable is a production newsletter platform shipping with Next.js 16, Prisma 7, Stripe, Auth.js, and twelve agents working from the same protocol library. We dogfood reveren before we ship it. The numbers below are real, not aspirational.
protocols shipped
all MIT in the base libraryagents tested
pipeline runners
tests passing
code coverage
data models
Six things we deliberately don't do.
Counter-positioning matters more than feature lists. Here's where we draw lines — including the one that separates an open marketplace from a vendor-locked one.
Not an AI agent.
We don't write code, run inference, or replace Claude / Cursor / Copilot / Windsurf. We make the agent you already use dramatically better.
Not an IDE.
Live in whichever editor you already love. We coordinate the agent inside it.
Not a model provider.
Bring your own model. We're the standards layer between your codebase and whichever model you pay for.
Not a knowledge base.
Notion, Confluence, internal docs already exist. reveren reads them; it doesn't replace them.
Not single-vendor lock-in.
The .protocols/ format is open. Your investment in protocols travels with you across agents, IDEs, even off-platform.
Not a vendor-locked marketplace.
Anthropic Skills runs on Claude. ChatGPT GPT Store runs on ChatGPT. The reveren Pod Marketplace runs on every agent your team already pays for. Author once. Sell on every agent.
Single-file rules vs. a real standards layer.
Vendor-specific instruction files solve a piece of the problem. They don't chain into pipelines, don't run from CI, don't survive a tool switch.
"The agent doesn't need a smarter model. It needs an operating manual."
The .protocols/ format is yours.
Every protocol is plain Markdown with a typed front-matter block. Version-controlled in your repo. Readable by humans, parseable by agents, portable between vendors.
We publish the spec under CC-BY 4.0. The reference CLI is MIT. If we disappear tomorrow, your team's protocols keep working.
"I built reveren because I was tired of correcting the agent. The fix isn't a smarter model. The fix is an operating manual the agent reads before it touches your repo. Try it for an hour. Tell me whether the output got better."
Innocent Muisha
Founder, reveren