whereismyai
AgentScore
AgentScore is an open-source AI agent monitoring platform that tracks production traces and automates fixes.
Latitude is an open-source AI agent monitoring and observability platform designed to track production traffic and uncover failures. It analyzes completed sessions to extract conversation patterns, groups recurring errors into prioritized signals, and alerts teams via Slack, email, or webhooks. When an issue appears, the platform dispatches coding agents like Claude Code or Cursor with context to implement fixes and open pull requests. It supports OpenTelemetry and SDK integration, custom signals, automatic golden datasets, and human annotations. The platform also offers advanced filters for semantic search, failure mode clustering, and an MCP server for workspace management.
AgentScore at a glance
- Best for: Software engineer
- Runs on: Web, API, Slack, self-hosted, open source
- Pricing: Starter Free; Pro $99/month; Enterprise Custom
- Released: September 2026
- Category: AI agent tool
- Website: latitude.so
Key features
- Open-source AI agent monitoring platform
- Production trace tracking and session analysis
- Automated failure grouping and prioritization
- Coding agent dispatch for automatic fixes
- OpenTelemetry and SDK integration support
- Automatic golden dataset generation
- Human annotations and inline feedback
- Advanced semantic search and filtering
Use it for
- Monitoring AI agent behavior in production
- Detecting and alerting on agent failures
- Dispatching coding agents to fix errors
- Building evaluation datasets from traces
Alternatives to AgentScore
- Composio — Equips AI agents with tools, managed user auth, and sandboxes across 1,500+ apps
- CodePilot — Consolidates multiple AI agents and local model connections in one workspace.
- Bedrock AgentCore — It eliminates the heavy lifting of managing specialized agent infrastructure.
- Polylane — It autonomously investigates production errors and submits ready-to-merge PR fixes.
- Tessary — Monitors every production trace and investigates root causes using trace and code evidence.
- Surfcore — Gives models a persistent identity, decaying memory, sleep schedule, and multi-platform presence.
- OpenViking — Unifies agent memory, knowledge RAG, and skills into a single directory structure.
- Alook — Connects existing local coding agents to team communication channels.
- Agent Beacon — Retains coding agent knowledge across tools so errors do not repeat.
- Agentic Context Engine — Automatically investigates agent errors and generates reviewable code fixes.
- AgentsView — It keeps your AI coding session archive local by default while tracking costs and activity.
- VoltAgent — Offers open-source flexibility with built-in supervisor agent orchestration and RAG.
All AgentScore alternatives