// AGENTS

Agents that deliver

Production-ready AI agents for real business workflows. Handle customer support, process data, generate content, and automate operations—reliably and at scale.

What Agents Can Do

Call Tools

Execute functions, make API calls, read files, run code, send emails—agents decide which tools to use and when.

Schedule Tasks

Break complex goals into steps. Create task queues. Execute autonomously until the goal is complete.

Remember Context

Persistent memory across sessions. Remember user preferences, past interactions, and learned information.

Self-Reflect

Evaluate outputs before returning. Catch errors. Improve responses. Learn from mistakes.

Built for Production

Enterprise workloads demand reliability. Apteva agents run in isolated processes with dedicated resources—when one fails, others continue unaffected.

Automatic health checks, graceful degradation, and instant recovery ensure your business operations never stop. Monitor everything with built-in observability.

Scale from a proof-of-concept to production without re-architecting. The same agents that work in development work at enterprise scale.

Agent Loop
while goal not complete:

  1. THINK
     └─ analyze current state

  2. PLAN
     └─ create/update task list

  3. ACT
     └─ call tools, execute tasks

  4. OBSERVE
     └─ check results, update memory

  5. REFLECT
     └─ evaluate, learn, improve

Built-in Agent Types

Worker Agent

Executes tasks and returns results

core

The default agent type. Receives a task, processes it using the configured LLM, and returns the result. Stateless between invocations.

Conversational Agent

Maintains context across messages

core

Persists conversation history in local storage. Supports threads, branching, and context windows. Ideal for chat interfaces and interactive sessions.

Supervisor Agent

Manages and coordinates other agents

core

Spawns, monitors, and restarts child agents. Implements supervision strategies like one-for-one and one-for-all. Handles failures gracefully.

Tool Agent

Executes functions and external calls

extended

Equipped with tool definitions and function calling capabilities. Can execute code, make API calls, read files, and interact with external systems.

Custom Agent

Define your own behavior

extended

Create agents with custom initialization, message handlers, and lifecycle hooks. Full control over the agent's behavior while leveraging the runtime.

Agent Lifecycle

1
spawn
2
init
3
running
4
stop
spawn

Process is created with configuration and assigned a unique ID.

init

Agent loads state, connects to LLM provider, registers with supervisor.

running

Agent processes messages, executes tasks, maintains state.

stop

Graceful shutdown, state persistence, resource cleanup.