Home / Platform
The Camaan platform
Build, run and govern AI agents. In one place.
Agents, durable workflows and repeatable scenarios on your own data, under rules you set per action, with every step signed into the audit ledger.
Agent builder, workflow engine, lake engine, ontology, knowledge graph, evaluations, model gateway, audit store and deployment, built to work together.
Runs where you need itOur cloud, your cloud or your own servers.
Architecture
The stack, layer by layer.
Products sit on the same data, rules and ledger as everything you build yourself. Governance runs across every layer. Go deeper: proof and audit, knowledge and ontology, the marketplace.
Assistants nobody owns, and nobody can explain.
Agents
Build agents from tools, knowledge and skills. Every save is a version you can compare, you test against real cases before anyone relies on it, and every action is recorded with who asked for it.
- Versioned on every save
- Evaluations before release
- Tools, skills and knowledge per agent
- Recorded with who asked and why
Automations fail halfway, and nobody notices.
Durable workflows
Workflows join agents, tools and people into one process, started by a schedule, an event or a request. Each step is saved as it completes, so a restart or a failed call doesn't lose the run.
- Every step saved as it completes
- Schedules, events and requests
- A person's approval wherever needed
- Inputs and outputs kept for every run
Ask an AI the same thing twice, get two answers.
Scenarios
A scenario pins down how a recurring task is done: the data, the steps and the rules. The agent fills in only what changes, so results repeat, can be checked, and use a fraction of the tokens.
- The same answer to the same question
- Results you can check
- Far fewer tokens per run
- Charts and outputs your teams already use
Agents query with more access than the person asking, and don't know what your terms mean.
Lake engine, ontology and knowledge graph
Connect databases, files and warehouses, map them into governed tables and follow lineage from source to answer. The ontology describes your business in your words; the knowledge graph fills it with your records.
- Federated queries or scheduled copies
- Row rules and masks inside the engine
- Lineage from source to answer
- Ontology Studio: build it by describing your business
No one can prove what the AI did, or who allowed it.
Governance and audit
For each kind of action you choose whether an agent acts alone, asks a person first, or never does it. Every action is signed into the ledger with its reason, the rule that decided it, and who approved it and when.
- Read, write, run code, spend, delete
- Approvals by a second person
- Signed records in plain language
- Identity, SSO and SCIM
A new platform usually means replacing what you run.
Runtimes
Use the models you already pay for, hosted or on your own servers. Run agents' code in the built-in engine or in your own cloud account, and connect the engines you already run.
- Any model provider, or your own
- Budgets and routing per model
- Code sandboxes
- Your cloud or your servers
Governed actions
What happens when an agent wants to act.
The same five steps for every action. What changes is the rule you set for its kind.
