Guide · 2026-06-11 · OpenAgent.bot Editors

Open-Source AI Agent Projects Found Today: June 11, 2026

Today's discovery pass adds projects across local model serving, asynchronous coding agents, agent memory, NotebookLM skills, MCP extraction, evaluations, and agent-team bots.

Today's open-source agent discovery pass found useful additions in every OpenAgent category. The pattern is practical: local model runtimes, asynchronous coding agents, sidecar memory, research skills, MCP extraction, evaluation platforms, and agent-team bot systems.

Today's additions

CategoryProjectWhy it matters
ModelsRapid-MLXApple Silicon local AI engine with OpenAI-compatible API and tool calling
AgentsOpen SWELangChain's open-source asynchronous coding agent
Memory systemsMnemo CortexMemory coprocessor for AI agents with persistent semantic recall
Skillsnotebooklm-pyUnofficial NotebookLM API and agentic skill for research workflows
PluginsAgentQL MCPMCP server for structured web data extraction via AgentQL
ToolsFuture AGISelf-hostable platform for evals, tracing, simulations, datasets, gateway, and guardrails
BotsWegentAI-native operating system for organizing and running agent teams

What stood out

The strongest projects today are infrastructure pieces rather than generic wrappers. Rapid-MLX improves the local model layer, Open SWE tackles asynchronous software work, Mnemo Cortex adds a memory coprocessor, and AgentQL MCP gives agents a cleaner web extraction tool surface.

How to evaluate this batch

For Rapid-MLX, test tool-call compatibility and latency on your own Apple Silicon device. For Open SWE, start with a low-risk repository task and require code review. For Mnemo Cortex, compare recall quality across repeated agent sessions. For notebooklm-py, remember that it is unofficial and validate compatibility. For AgentQL MCP, begin with non-sensitive web pages. For Future AGI, instrument one agent path before adding dashboards. For Wegent, test a small agent team before expanding roles.

Official sources