The work it removes is integration: a question about a wallet's history and the sentiment around a token would otherwise mean separate accounts, separate keys and separate billing at each provider. Planning being deterministic matters more than it sounds — it means the same question is answered by the same sources twice running.
A routing layer in front of paid data providers. In its default mode a request is planned and executed in one call; planning is deterministic, so the same question composes the same set of provider calls.
- One endpoint and one key covering every provider behind the gateway
- Several provider APIs composed inside a single request
- Deterministic planning — semantically identical prompts produce the same composition
- A separate planning step, when you switch it on, so the plan can be seen before it runs
- The crypto vertical at launch: market intelligence, wallet transactions and relations, whale movement and sentiment, from providers including Dune, DefiLlama, Zerion, Xverse, Coingecko and Cielo
A Kirha access code — the beta is invite-based, and you request one from the Kirha app. The separate planning mode is off unless you turn it on with an environment switch.
One command plus a key — npx @kirha/mcp-installer install --client <client> --key <api-key>, then supply credentials