Galdor adds local tracing and replay for Go AI agents
Galdor is an for building and running AI agents in Go. Version 1.0.0 was released in June 2026 under the Apache 2.0 license. Its main focus is making agent behavior easier to inspect: , tool calls, and workflow steps can be traced and viewed through a local dashboard backed by SQLite.
The dashboard runs from the same binary, so teams do not need a separate paid monitoring service just to see what happened during an agent run. Galdor can record a real run and replay it later under controlled conditions, which helps debug failures and compare changes before shipping them. It supports several s, including OpenAI, Anthropic, , , Ollama, and vLLM-style setups.
It also includes MCP and A2A support on both the client and server side, built-in patterns, short-term and options, and an evaluation framework.
Key points
- Galdor is a Go framework for building AI agents, released as v1.0.0 in June 2026.
- It includes local tracing, a SQLite trace store, and a built-in dashboard.
- Replay lets teams rerun recorded agent behavior to debug and compare changes.
- Multiple s are supported, which can help test cheaper model choices.
- MCP, A2A, patterns, memory, and evaluation tools are included.