Kohala Platform — build, govern & deploy your own AI agents
Build and govern your own AI agents on Kohala's framework-agnostic runtime. Author in LangGraph, CrewAI, OpenAI Agents SDK, and more — per-run governance, audit trail, memory API, spend caps, and metering built in.
Frequently Asked Questions
- What is the Kohala platform?
- Kohala is the governance, memory, and monetization fabric for AI agents. You author an agent in any framework — LangGraph, CrewAI, the OpenAI Agents SDK, or Kohala's own typed SDK — and run it on a governed runtime that enforces a charter, tool allowlist, validators, spend caps, and an audit trace on every run. It provides a portable tenant-scoped memory API, built-in metering, and a marketplace to publish and sell what your agents produce.
- Do I have to use a specific agent framework?
- No. Kohala is framework-agnostic. Bring the agent loop you already wrote in LangChain, LangGraph, CrewAI, PydanticAI, the OpenAI Agents SDK, Mastra, the Vercel AI SDK, n8n, or Vellum — or use Kohala's first-party SDK. Adapters map your agent onto the same governed runtime, memory, and metering. The governance model is the constant; the authoring tool is your choice.
- What does governance-per-run mean in practice?
- Every agent run is checked against a charter (a plain-language statement of what the agent may and may not do), a tool allowlist (the exact skills it can call), output validators (checks every run must pass before results ship), a spend cap (a hard ceiling on credits and cost per run and per period), and an audit trace (a tamper-evident record with trace ID, latency, and credit cost). These are enforced on every run whether your agent loop runs on Kohala's runtime or your own infrastructure.
- How does Kohala meter agent costs?
- Every agent run is metered in platform credits against a per-agent billing cap you set. The run reports its exact credit cost, latency, and a trace ID, so you always know what each execution cost. Enterprise accounts can bring their own LLM provider keys and get billing waivers on those tiers; spend caps are enforced independently and can never be exceeded.
- Can I start for free without hosting?
- Yes. Use the open-source devkit — a CLI, a local emulator, and an open MCP memory server — to build and run the whole agent loop free on your own machine. When you're ready, one command deploys the same agent to Kohala's governed, metered, monetizable runtime. Personal accounts include 1,000 credits at registration, then pay-as-you-go top-ups.
- How does Kohala support enterprise and developer requirements?
- Enterprise accounts add bring-your-own data tiers (S3, Pinecone, LLM provider keys), SSO, SCIM provisioning, full audit logs exportable for compliance, and account-scoped MCP fleet control. Developers can use the open SDK and devkit to author in any framework, call the REST API or webhooks, and publish agents and Koans to the marketplace.
- Can I use Kohala from a mobile app?
- Yes. Business and Kohala Enterprise accounts can mint mobile application tokens — long-lived, cookie-free bearer tokens for native apps. A mobile token can chat with Lani (including attaching documents and photos that Lani reads and discusses), act on Canvas activity and approvals, view reports and Koans, and check the account's credit balance, all with the same tenant scoping and governance as the web app. See the API reference for the full mobile surface.