# Honeypoard

> Connect the systems you already use. Turn their data and capabilities into useful views, grounded answers, and approved actions for people and assistants.

Website: https://honeypoard.com/
Product guide: https://honeypoard.com/product-guide

Honeypoard connects REST APIs, GraphQL endpoints, and internal services. Optional connectors can bring in common SaaS tools. A company can use those connections in its Honeypoard workspaces, publish views inside its product, or offer governed capabilities to coding harnesses through MCP and custom agents through A2A.

## How Honeypoard works

Honeypoard turns connected operations and data into named capabilities, such as finding an order, explaining an invoice, or requesting a refund. A capability can combine several services. Honeypoard identifies candidate entities and relationships between sources, including join keys and tenant boundaries, and shows the evidence and uncertainty for review. Its capability brain can suggest mappings and useful capabilities, but a suggestion is not automatically published.

The approved model is represented by versioned, reviewable definitions. `capabilities.yaml` describes source bindings, entities, relationships, inputs, outputs, and execution plans. `governance.yaml` describes audiences, permissions, field restrictions, limits, and approval rules. Honeypoard retains the approved configuration and revision history. Source changes create proposed revisions instead of silently changing published behavior.

After review, Honeypoard validates and compiles the definitions into a pinned, predictable capability contract. Published tools follow the approved execution plan and enforce governance at discovery and invocation. An LLM may help propose a relationship or explain an answer, but it does not decide access or rewrite a published tool during a call.

## Where capabilities can be used

- **See:** Build private, shared, or embedded views from approved capabilities, with source context and freshness.
- **Ask:** Ask questions across permitted sources and trace answers to their inputs.
- **Act:** Publish selected capabilities for assistants and agents. Coding harnesses can access a user's authorized workspace or workspaces through hosted MCP. A company's custom agents can connect to its workspace through A2A. Each path exposes only what the caller is allowed to use, and consequential actions can require human approval.

Both MCP and A2A use the published capabilities and enforce workspace permissions, capability rules, and approval requirements.

## Data and hosting

Honeypoard Cloud stores workspace configuration, approved definitions, revisions, and publication state. A separate data plane holds connector credentials, runs capabilities, and handles business records and results. Business data stays in connected systems unless a capability uses an approved sync or cache in the data plane. Honeypoard manages the data plane by default; Enterprise customers can host it in their own environment.

## Explore the product

- [Product guide](https://honeypoard.com/product-guide): Read how connections become reviewed capabilities, where governance applies, and how deployment boundaries work.
- [Governance](https://honeypoard.com/#control): Choose approved sources and audiences, set access rules, and require a person to review actions above a chosen limit.
- [Inside Honeypoard](https://honeypoard.com/#product): See the product experience for dashboards, questions with source context, agent access, and approvals.
- [See, Ask, Act](https://honeypoard.com/#possibilities): Build private or embedded views; ask across approved sources and trace the answer; give harnesses and custom agents approved capabilities through MCP and A2A.
- [Architecture](https://honeypoard.com/#deployment): Use Honeypoard Cloud or run the data plane in your own environment. Both support the same views, answers, and governed agent capabilities.

## Request access

Tell us what you want to connect and build through the [contact form](https://honeypoard.com/#contact).