Managed tool access
Enable platform and custom tools at the domain level, then expose only what each agent-enabled intent needs.
Platform
Agent orchestration
Give intents tools without giving up control.
Turn on agent behavior for the capabilities that need multi-step work. LiyaEngine coordinates tool access, intermediate decisions, and execution traces inside the intent boundary.
Agent run
Ready
PlanToolsMemoryTrace
Live configuration
1 retrieve_policy
2 calculate_exposure
3 call_approval_api
4 compose_response
2 calculate_exposure
3 call_approval_api
4 compose_response
Tools used3
GuardrailsPassed
ResultCompleted
Intent-scopedAgent behavior serves a defined capability.
Tool-controlledOnly approved tools are available to a run.
TraceableIntermediate work is visible after execution.
Platform capabilities
Everything needed to operate this layer.
Build the capability once, keep its operating context visible, and improve it without spreading AI infrastructure across your application.
Multi-step execution
Coordinate retrieval, calculations, and API calls before producing a final response.
Context continuity
Carry the relevant state through a run without turning the application into an orchestration engine.
Failure handling
Surface tool and model failures as inspectable run states instead of opaque response errors.
Policy checks
Apply guardrails around tool inputs, outputs, and the final answer path.
Run traces
Review what the agent attempted, which tools ran, and how the final response was assembled.
Operating workflow
From configuration to production evidence.
The interface follows the work in the order teams actually perform it.
01
Choose the intent
Start with a capability that genuinely needs tools.
02
Grant tool access
Enable only the functions required for the job.
03
Test the run
Inspect intermediate steps and failure paths.
04
Operate with traces
Use run history to improve behavior over time.
What improves
A cleaner operating model for production AI.
Move complexity into a visible platform boundary while keeping the product integration direct.
01
Bounded autonomy
Agent behavior stays attached to a named purpose and an explicit tool set.
02
Simpler application code
Orchestration logic runs in the platform instead of being duplicated across services.
03
Operational clarity
Teams can debug tool use and model decisions from one execution record.
Questions
Before you build.
Practical answers about how this part of LiyaEngine fits into the platform.
Do all intents need Agent Mode?+
No. Standard intents are the simpler choice for direct structured responses. Enable agent behavior only when the task needs tools or multiple steps.
Where are tools configured?+
Platform and custom tools are enabled in the domain, then used by agent-enabled intents within that domain.
Are agent runs observable?+
Yes. Run traces show intermediate steps, tool activity, and the final outcome.
Build the capability. Keep the operational context.
Start with a tenant-owned domain and expose the first intent through a stable, governed API.