Open Mercato Review
Open-source TypeScript foundation for AI-built commerce, CRM, and ERP
Agent Skills · Open Source · OPEN SOURCE
Overview
Open Mercato is an MIT-licensed TypeScript framework that positions itself as an AI-engineering foundation for commerce, CRM, and ERP builds: business modules and platform conventions ship pre-decided, so human developers and AI coding agents build features instead of re-arguing architecture on every prompt. The pitch is starting at 80% done. The stack is a Yarn-workspaces monorepo on Next.js App Router with MikroORM, zod, Redis, and Meilisearch, and the core package ships catalog, sales (quoting, ordering, fulfillment, billing, payment gateways, shipping carriers), a CRM module with people, companies, deals, and activities, a customer self-service portal, checkout, workflows, business rules, custom entities, translations, dashboards, audit logs, and API keys. Multi-tenancy is the default posture: a directory module provisions tenants and organization trees, tenant and organization context propagates so queries filter themselves, and per-tenant field-level encryption uses AES-GCM. RBAC is two-layer, roles bundling module.action feature strings with per-user, per-tenant overrides. The AI story is documented rather than implied: the repo carries an AGENTS.md spec-first workflow, a standalone AI development harness with 192 evaluation cases and tool configuration for Claude Code, Codex, and Cursor, an in-product AI framework with typed module agents and a mutation-approval gate, and an MCP server exposing about 70 tools. An Enterprise Edition package adds SSO with OIDC and SCIM, MFA, and record locks outside the MIT core, with no published pricing. A public demo runs at demo.openmercato.com, sandboxes at sandboxes.openmercato.com, and community chat on Discord. Version 0.7.0 shipped on August 26, 2026, and the requirements are honest about weight: Node 24, PostgreSQL 17 with pgvector, Redis 7, and Meilisearch via Docker Compose. This is a foundation framework, not a product: compare it against starting a Next.js commerce project from scratch, not against Shopify or a configured CRM.
AI Capabilities
- AI development harness with 192 evaluation cases
- MCP server exposing about 70 tools
- LLM email triage with human approval gate
How to install
- Monorepo quick start from the docs: git clone https://github.com/open-mercato/open-mercato.git, cd open-mercato && git checkout develop, docker compose up -d, then cp apps/mercato/.env.example apps/mercato/.env and set DATABASE_URL, JWT_SECRET, and REDIS_URL.
- Continue with yarn install, yarn build:packages, yarn generate, then yarn build:packages a second time (the docs require it), then yarn initialize, which runs migrations, seeds roles, provisions an admin, and loads demo CRM data (pass --no-examples to skip demo data).
- Start development with yarn dev; the backend app runs at http://localhost:3000/backend with credentials printed by yarn initialize, and the splash page on port 4000.
- For a standalone project rather than the monorepo: npx create-mercato-app my-app. The CLI binary is yarn mercato, used for commands such as auth setup, db:migrate, entities install, and api_keys add.
- Upgrades follow git pull && yarn install && yarn db:migrate && yarn generate && yarn dev. Development variants include yarn dev:greenfield, dev:ephemeral, dev:classic, and dev:verbose.
Requirements
Node.js 24 with Yarn 4 via corepack. Docker Compose brings PostgreSQL 17 with the pgvector extension, Redis 7, and Meilisearch (ports 5432, 6379, and 7700). The MCP server runs separately (yarn mcp:dev on port 3001, yarn mcp:serve in production).
Best for
TypeScript teams standing up custom commerce, marketplace, or CRM backends with an AI-assisted development workflow, who want multi-tenancy, RBAC, and business modules already decided so agents and developers spend their time on the differentiating features.
Not for
Teams that want a CRM or storefront they configure rather than code, and anyone needing a turnkey admin experience on day one. Enterprise features (OIDC/SCIM SSO, MFA, record locks) live in a separate package with no published pricing, so regulated buyers need to talk to the vendor first.
Hands-on notes
Assessed from the repository, docs.openmercato.com, and openmercato.com; we have not run it. The project is young (repo created September 2025) and moving fast: v0.7.0 on August 26, 2026 added the standalone AI development harness, an EUDR module, a warranty and RMA desk, and mobile push, following v0.6.7 with a phase-one WMS. Expect breaking change cadence typical of 0.x software.
The AI tooling is the substance behind the marketing framing, and it is unusually concrete for an open-source project: an AGENTS.md with a 32,768-byte budget and Always/Ask-First/Never rules, spec-first design in .ai/specs/, per-agent configuration for Codex, Claude Code, and Cursor, 192 evaluation cases, and a documented MCP server with about 70 tools plus a full Claude Code setup guide. CLAUDE.md in the repo is a one-line pointer to AGENTS.md.
Two documentation gaps are worth knowing before you commit: the website still claims version 0.4.6 against an actual v0.7.0, and the docs' pinned Yarn version (4.12.0) disagrees with package.json (4.17.1). Neither is serious, but both mean you should trust package.json and GitHub releases over the marketing site.
Scale expectations matter: 1,700-plus stars, a public Discord, a demo at demo.openmercato.com, and sandboxes at sandboxes.openmercato.com signal an active early project, not an established platform. The roadmap (visual module scaffolder, GraphQL gateway, multi-region tenancy, plugin marketplace) is documented but unbuilt.
Verdict
A credible AI-first foundation for engineering-led commerce and CRM builds; early, fast-moving, and not a turnkey product.
Pros and cons
| Pros | Cons |
|---|---|
| ✓ Open-source licensing with free self-hosting | |
| ✓ API access for custom integrations | |
| ✓ AI capabilities: AI development harness with 192 evaluation cases | |
| ✓ Established community (1,708 GitHub stars) |
Building your martech shortlist?
The weekly newsletter: one tool teardown, one workflow, no fluff. Free.
Frequently asked questions
What is Open Mercato?
Open-source TypeScript foundation for AI-built commerce, CRM, and ERP. It ships with AI development harness with 192 evaluation cases, 1,708 GitHub stars. MartechSignal's review covers features, pricing, and how it compares to alternatives.
How much does Open Mercato cost?
Open Mercato is open source and free to self-host. Hosted plans may add support and managed features.
Is Open Mercato a good agent skills tool in 2026?
A credible AI-first foundation for engineering-led commerce and CRM builds; early, fast-moving, and not a turnkey product.
Is Open Mercato a ready-to-use CRM or ERP?
No, and the project says so itself: it is a foundation framework, with the pitch that business modules and conventions are pre-decided so you start at 80% and build the differentiating 20%. The core CRM module does ship with people, companies, deals, and activities, plus a customer self-service portal, and a demo with sample CRM data loads during yarn initialize, so you can see working software at demo.openmercato.com. But the intended comparison is against starting a Next.js commerce project from scratch, not against Shopify or a configured CRM, and evaluation should assume engineering work.
How does Open Mercato work with Claude Code, Codex, or Cursor?
The framework is built around agent tooling. The repo's AGENTS.md defines a spec-first workflow (designs live in .ai/specs/ as dated markdown files) with Always, Ask-First, and Never rules, and CLAUDE.md simply points at it. The standalone project generator emits an AI development harness with guides, skills, and tool-specific configuration for Codex, Claude Code, and Cursor, backed by 192 evaluation cases and a sandboxed release gate. At runtime, a documented MCP server exposes about 70 tools for agent access, and shared skills install with npx skills add open-mercato/skills. In-product, an AI framework provides typed module agents with a mutation-approval gate before AI-driven changes land.
What does Open Mercato cost to run?
The core is MIT-licensed and free to self-host, including all documented core modules, so costs are your infrastructure (PostgreSQL 17 with pgvector, Redis 7, Meilisearch) and engineering time. A separate Enterprise Edition package adds SSO with OIDC and SCIM 2.0, MFA, sudo re-authentication, and record locks, and ships outside the MIT core; no price is published anywhere on the site or docs. The site mentions community and commercial support tiers without amounts, so enterprise pricing and support are quote-based by absence of a published list.