ComponecatAI software catalog: built for agents and your team
Componecat is a software catalog built to be read by AI agents as well as by your team. It is not a catalog of AI models: it is the registry of your own systems, services and libraries, structured so an agent can query it over MCP and get exact answers.
What makes a catalog readable by AI
An agent cannot ask a colleague what a field means. It needs the catalog to be explicit:
- Typed structure. Every entity has a kind, and the kind decides its fields and features: a repository link, documentation, interfaces, an activity log, children. Agents get fields, not free text to interpret.
- Typed relationships. Each relationship is directed and has a kind, such
as
parent-ofordepends-on, and you mark which kinds count as dependencies so impact analysis follows them. - Search by meaning. On Pro and Enterprise, with an embedding model
configured, a search can rank by meaning
(
~"the service that sends invoices"), and agents get tools that take a plain-language description directly. - Served over MCP. The whole catalog is available to any MCP client, with access scoped per resource. See the MCP server.
- Kept current from your tools. On Pro and Enterprise, the catalog reads descriptors and Markdown documentation from your repositories, and syncs team ownership from GitHub, GitLab, Atlassian or Microsoft 365.
Your taxonomy, not ours
Componecat starts you with built-in kinds and relationship kinds, and you add your own: a kind for data pipelines, a field for PCI scope, a relationship kind for "publishes to". The descriptor format teams commit uses the same kinds and fields, documented in the descriptor reference.
Built for your team too
The same catalog is what your engineers browse:
- One catalog page with a hierarchy tree, a ranked list and a relationship graph, full-text search, filters and a query language.
- Documentation, interfaces and an activity log on each entity whose kind turns them on.
- Scorecards that check every entity in scope against your standards on a schedule.
- A wiki for the conventions that apply across entities. Agents are told to read it before they create or edit one.
- An AI assistant in the app that answers from the catalog, on a model you choose: OpenAI, Azure OpenAI, Anthropic, AWS Bedrock or Ollama.
Plans
Free covers up to 5 users and 50 catalog entities. Basic is $20 a month for up to 20 users and 250 entities, and adds the REST API, plus import and export. Pro is $100 a month for up to 100 users and 2,000 entities, and adds search by meaning, integrations, outgoing webhooks, service accounts and single sign-on. Enterprise has no entity limit and custom user limits. Compare them on pricing.