Back to Ecosystem Pulse
ProtocolExperimentalMCP
Mimir
by orneryd
Open memory bank + vector search with drag-and-drop multi-agent orchestration
Go
Updated Dec 25, 2025
Share:
Overview
Provides a fully open, customizable memory bank with semantic vector search for locally indexed files and shared session memories. Combines code intelligence (file indexing and embeddings) with a drag-and-drop multi-agent orchestration UI so worker agents can persist lessons and learn from past errors. Includes vector search, graph embeddings, and session-shared memories that surface agent history across chat contexts.
Why It Matters
As agents run repeatedly and delegate work, persistent memory and searchable traces let teams spot recurring failures and build agent track records. Mimir makes agent mistakes and corrective actions discoverable by storing indexed artifacts and memories across sessions, enabling trust signals and continuous learning. That visibility is essential for [multi-agent trust] and practical A2A evaluation workflows.
Ideal For
Teams building multi-agent systems that need persistent, searchable memories and a UI to inspect agent runs and learn from past failures.
Applications
- Persisting agent decisions and error traces to build an agent track record
- Indexing code and developer artifacts for semantic search during agent runs
- Inspecting and replaying multi-agent workflows via a drag-and-drop UI to debug failure modes
Topics
code-intelligencecodebasedockergraph-algorithmsgraph-apigraph-databasegraph-embeddingindexingllm-orchestrationllmops+9 more
Similar Tools
langchainautogen
Keywords
multi-agent orchestrationsemantic-searchagent-memorymulti-agent trust