Capability Planning: help your assistant know what is missing
Private Beta available by request: request a demo.
Capability Planning gives an MCP client a way to ask a useful question before it acts: “Given what we know, what can happen next?”
The answer is not a vague recommendation. It is a bounded proposal built from the business states and actions that an API team has explicitly described.
Four clear answers
For a requested goal, the assistant can receive one of four plain-language outcomes:
| Outcome | What the assistant can tell the person |
|---|---|
| Complete | “This is already done. There is nothing else to do.” |
| Planned | “Everything is ready. Here is the safe next step.” |
| Clarification required | “I can help, but I need this specific detail or confirmation first.” |
| Unplannable | “There is no declared, safe route to that result from the information available.” |
Planning never calls the underlying API. It does not book, cancel, charge, update, or delete anything. It prepares the conversation so a person and an MCP client can make the next decision with their eyes open.
Dynamic, not rigid
This is not a prewritten chatbot flow. The assistant can respond naturally to what a person has already said or selected. The plan changes with the real situation:
- A provider was found, but the person has not selected one yet.
- A time was selected, but the patient details still need verification.
- Everything is ready, but the booking needs confirmation.
- The desired appointment is already booked.
The assistant does not have to restart a script. It can ask the one question that matters now.
Where Jev fits
Capability Planning itself remains deterministic: it follows the explicit business states, authorization, and safety gates defined for the API.
In the v1.2 Private Beta, the optional Jev-powered TypeSafe classifier can help earlier in the experience: interpreting a person’s intent, handling ambiguous wording, and ranking eligible candidate tools. Jev supplies a structured semantic judgment; HAPI still decides what is permitted and what requires confirmation.
That combination gives teams a useful balance: conversational intelligence where language is fuzzy, and predictable safeguards where the business process must be exact.
A safe path to adoption
- Start with Capability Graph to improve tool discovery.
- Describe a small number of high-value, well-understood business transitions.
- Test the four planning outcomes with realistic scenarios.
- Let the assistant prepare plans before considering any controlled execution.
Want help identifying the first workflow that is a good fit? Request a Private Beta demo.