Give your AI a better sense of what comes next
Private Beta available by request: request a demo.
An MCP server can expose a lot of useful tools. But a helpful AI assistant needs more than a menu. It needs to know what the person is trying to achieve, what information is already available, what is missing, and whether a change needs a human decision.
HAPI Capability Planning turns an API into a practical decision guide for AI clients. It helps an assistant move a conversation forward without guessing its way through an important workflow.
“I can do that. First, I need you to choose a provider and a time.”
“Everything is ready. This will cancel the appointment. Would you like me to continue?”
That is the promise: less improvisation, clearer conversations, and safer automation around the systems your business already runs.
What is coming in HAPI v1.2?
HAPI v1.2 brings together three complementary capabilities:
| Capability | What it helps an MCP client do |
|---|---|
| Capability Graph | Find the most relevant tools in a large API without presenting every possible operation at once. |
| Capability Planning | Understand whether a goal is already complete, which known detail is missing, or what safe step could happen next. |
| Jev-powered classification | Add optional semantic judgment to shortlist and rank the right tools when human language is nuanced or ambiguous. |
The foundation stays deterministic. HAPI applies authorization and declared workflow rules before a classifier is asked for a semantic judgment. Planning prepares a proposal; it does not make an API call or change data.
Why Jev?
The optional classifier is powered by Jev from TypeSafe, a System One model designed for fast, structured decisions. Rather than asking a generative model to improvise an entire workflow, HAPI can give Jev a bounded set of eligible tools and ask it to help choose or rank the best fit for what a person means.
That makes Jev a natural complement to an MCP server:
- HAPI keeps the API contract, authorization, and safety rules in charge.
- Capability Graph narrows the available options deterministically.
- Jev adds semantic judgment when words can mean more than one thing.
- The assistant can ask a focused question when confidence or context is not enough to choose safely.
Learn more about Jev and TypeSafe System One models.
See it in action: appointment scheduling
Someone says, “I need to see a pediatrician next week.”
The assistant can guide the person through the real process:
- Find suitable, verified professionals.
- Ask the person to select one.
- Check that professional’s live availability.
- Ask the person to choose an available time.
- Verify the patient details.
- Ask for confirmation before booking.
If the person wants to cancel instead, the assistant follows a different path: identify the appointment, verify the person, and ask for clear confirmation before a change that cannot be undone.
The conversation stays natural. The business process stays clear.
Start with the outcome you need
- Want an assistant to find the right tool from a large API? Start with Capability Graph.
- Want the assistant to explain what is missing and prepare the next safe step? Read Capability Planning.
- Interested in the v1.2 Private Beta and the optional Jev classifier? See Private Beta or request a demo.
Frequently asked questions
Is this another chatbot builder?
No. HAPI works with the OpenAPI and MCP tools you already own. Capability Graph and Planning give AI clients the context to use those tools more thoughtfully.
Does Jev control my API or make changes by itself?
No. Jev is an optional classifier for understanding and ranking intent among eligible choices. HAPI keeps authorization and declared safety requirements in the runtime. Planning is read-only, and execution remains a separate, explicit decision.
Does Capability Planning force every conversation into a script?
No. It helps the assistant react to the current situation. It can identify a missing choice, propose the next useful action, or recognize that there is nothing left to do.