Agentloop

Agentloop is a Go library that handles the execution loop, budgets, and hooks required for building AI agents.

Agentloop is a Go library that takes care of the core execution loop every AI agent needs, handling prompt sending, tool calls, and result aggregation until a final answer is produced. It includes budgets, hooks, skills, and tracing so developers can build custom agents without worrying about boilerplate scaffolding. The library integrates with llmhub to connect to providers like OpenAI, Anthropic, Gemini, Ollama, xAI, OpenRouter, and Z.AI without direct vendor API calls. It features two pre-configured reference agents for search and shell command execution, along with extensible hook structures for logging, tracing, and approval gates. Designed for developers building lightweight Go applications, it compiles into a static binary with a tiny memory footprint and no external runtime dependencies.

Agentloop at a glance

Best for: Software engineer
Runs on: CLI, self-hosted
Pricing: Free
Released: October 2026
Category: AI agent tool
Website: github.com/smhanov/agentloop

Key features

Provider-agnostic LLM integration via llmhub
Observable and configurable execution loop
Comprehensive hook system for logging and tracing
Built-in safety budgets and limit controls
On-demand skill loading with progressive disclosure
Detailed execution traces with token and cost usage
Static compilation with a tiny memory footprint

Use it for

Building multi-step web search agents with citations
Creating command-executing shell agents
Embedding AI agents into CLI tools or cron jobs
Developing custom conversational AI services in Go

Alternatives to Agentloop

AgentOps — Track token usage, costs, and traces across 400 plus LLMs and agent frameworks.
Composio — Equips AI agents with tools, managed user auth, and sandboxes across 1,500+ apps
Alook — Connects existing local coding agents to team communication channels.
Agent Squad — Runs multi-agent orchestration in Python, TypeScript, and locally on Apple devices.
Golem — Provides type-safe agent abstractions and zero-dependency Go implementation.
Loopy — Provides repeatable prompts with checks, proof, and stopping conditions.
Devloops — Adversarial loops let agents handle routine maintenance while humans make key decisions.
Search2o — Builds model-neutral agents with strict spend limits and role-based access.
Rakazo — Complete ownership of data, models, and execution environment.
Beeuni — Agents retain context and improve skills over time through a controlled review process.
AgentiLoop — A native Mac agent that works with 23 model providers or local models
MLflow — Complete observability and evaluation for production AI apps.

All Agentloop alternatives