Skip to main content

Understand Touba

Responsibilities written in code and enforced during execution

Touba defines the boundaries within which your agents act. Your applications and handlers retain their business logic; the language makes capabilities, permissions and approvals explicit.

Agents as Code

Describing an agent in a program gives your team a framework it can review, version and test: its mandate, accessible capabilities, permitted calls and required human decisions.

Declare

Write agents, skills and strategies. Specify the resources, budgets and approvals they require.

Compile

The compiler checks the program and produces AIR, the intermediate representation that preserves its contracts.

Execute

The runtime checks permissions before calls. JavaScript or Python handlers carry out business operations.

Inspect

The trace links decisions, calls and results to the source program. You can understand execution and audit its controls.

Explore Agents as Code

From program to your systems

CompilationYour programAgents · capabilities · limitsCompiled contractAIRAI modelProposes an actionTouba runtimeChecks before executionYour handlersJavaScript · PythonYour systemsAPIs · databases · servicesExecution traceDecisions · calls · resultsAuthorized callYour programAgents · capabilities · limitsCompiled contractAIRTouba runtimeChecks before executionYour handlersJavaScript · PythonYour systemsAPIs · databases · servicesExecution traceDecisions · calls · resultsAI model : proposes an action
The model proposes. Touba checks the call. Your handlers perform the action and the trace records its execution.

Integrate Touba into your application

Use the SDK for your environment to supply handlers, configure models and follow executions. Both SDKs share the same concepts and Touba contracts.

Create your first programExplore examples