OpenProse

Engineer your agents.

OpenProse brings contracts, composition, and version control to the Markdown your agents run on.

Talk to Us
Open source

Declarative contracts for the AI session.

A Prose program is a Markdown file with a contract — requires, ensures, strategies — not a chain of prompts. Forme auto-wires services by semantic match; any Prose Complete harness (Claude Code, Codex CLI, …) executes it. The program declares what, not how, so it gets better as models improve. All MIT.

1,000+ GitHub stars·7,000+ installs·Backed by Y Combinator·MIT licensed
View on GitHub →Read the spec →
---
name: hunter
kind: program
services: [analyst, ranker, compiler]
---

requires:
  - data_warehouse_url
  - codebase_ref

ensures:
  - brief: weekly growth findings, ranked by
    confidence × impact, each with an owner and
    a trivial next step

strategies:
  - prefer findings with code-level evidence
  - surface instrumentation gaps as first-class findings
In the wild

What people are saying

Sample deliverables

What a morning brief looks like.

Illustrative examples of the kinds of outputs programs ship into your Slack.

# growth-opsSave
H
hunterAPP6:00 AM

Weekly Growth Brief — Apr 7–13

Headline: a one-line config change is costing ~1,500 orders/day — and the fix is a one-line revert.

An engineer raised the grocery minimum from $10 to $15. No A/B wrapper, no tracking on threshold-blocked carts. Grocery conversion dropped 8.2% → 7.2% within 24h. Net impact: $45K–$67K in daily GMV.

#FindingImpactEffort
1Revert grocery min ($15 → $10)1.2–1.8K orders/dayTrivial
2Premium-tier prompts for high-freq users1.5–3K orders/daySmall
3Restore search card analyticsUnblocks measurementSmall

Full breakdown with queries and dashboards in thread.

Surfaces the one-line revert and the instrumentation gap that together represent seven figures of recoverable GMV the team couldn't see on its own.

# intakeGrow
T
intake-triageAPP7:00 AM

Intake Triage — 2026-04-13 · Pipeline Health: YELLOW

Five eligible participants ready for advancement. One safety flag requiring immediate attention. Two protocol questions for PI.

IDResultNotes
QR-…001EligibleY-BOCS 26, 2 adequate SSRI trials. Local.
QR-…003Not eligibleSAFETY FLAG — active suicidal ideation.
QR-…007EligibleHIGH PRIORITY — Y-BOCS 35, 3 failed trials.

If you do one thing today: schedule a follow-up with QR-…007 and begin consent. Most severe treatment-resistant profile in the batch — every hour of delay increases disengagement risk.

Turns a messy stream of survey and email inquiries into a prioritized, safety-checked enrollment queue the coordinator can act on before lunch.

Built by the builders

OpenProse runs on OpenProse.

Our marketing, GTM, and ops run on Prose programs. Thirteen responsibilities run the company — mention tracking, competitor intelligence, lead enrichment, outreach drafting, market indices.

See how we run our company →
mention-intelligencecompetitor-intelligenceopportunity-discoveryengagement-draftingagent-ecosystem-indexsaas-readiness-indexagents-site-analyticsperson-enrichmentcompany-enrichmentlead-enrichmentprogram-designercustomer-repo-scaffolderstargazer-intake
FAQ

Common questions

What is OpenProse?

OpenProse is a service-as-software company. You pay us a flat monthly fee and work gets done — reports generated, leads enriched, customer outreach drafted, operational briefs delivered — without anyone on your team opening a dashboard. The work runs on Prose programs, our open-source language for AI sessions.

How is this different from ChatGPT or Claude?

ChatGPT and Claude are tools you use. OpenProse is a service you receive. Your team doesn't prompt, doesn't configure, doesn't babysit. Outcomes arrive in your Slack, email, or the systems you already use — on a schedule, with human review where it matters.

How is this different from hiring someone?

Hiring takes weeks to source, months to ramp, and commits you to a full role even when you only need three specific responsibilities handled. OpenProse takes those responsibilities off your plate in days, runs them on a schedule with measurable confidence, and prices transparently — no salary band, no severance, no PTO.

What does a pilot look like?

We build one or two programs tailored to your team, run them for you, and deliver the output to Slack or email. Flat monthly fee, roughly $2,000 for two programs during pilot.

What if I want to run this in-house later?

You can. Prose is MIT-licensed and your programs live in a repo we share with you. Run them locally, fork them, hand them to someone else. The lock-in is quality of operation, not proprietary infrastructure.

What about mistakes — who reviews the output?

Every output passes through an automated review model before delivery. For work that goes externally to your customers, suppliers, or partners, we configure a human review step as part of the program.

How do you handle our data?

Your data stays in your org by default. OpenProse learns from what works across the platform — which program structures succeed, which patterns scale — but customer data never leaves your walls. Full tenant isolation is available for customers whose legal teams require it.

Who is OpenProse for?

Executives and operators at companies where the same operational work gets redone every week. Sales, marketing, support, research, product — any function with a repeatable pain that today gets handled by a person or gets dropped. We sell to the P&L owner, not to the individual whose job gets easier.

How is this different from LangChain, CrewAI, or AutoGen?

Those are libraries developers use to build agent systems. OpenProse is a company that delivers finished work product. We use our own open-source stack (Prose, Forme, Press) to run your programs; you never see the plumbing.

Can I see how you run your own company?

Yes — OpenProse runs on OpenProse. Our marketing, GTM, enrichment, and ops functions are Prose programs. The dogfood is the clearest picture of how the platform works in practice.