# MuBit

> A durable memory layer for AI agents — persistent across runs, queryable across sessions, with an explicit learning loop.

## Docs

- [Changelog Overview](/changelog): User-visible MuBit documentation and integration updates.
- [Concepts](/concepts): How MuBit's memory model works, what the learning loop does, and how it compares to vector DBs, prompt caches, and RAG.
- [Quickstart](/getting-started): Make your first authenticated MuBit call in under 60 seconds. The terminal on the right runs the same code — a real round-trip against `api.mubit.ai` showing every payload the SDK returns. Switch the terminal between Python, Node, and Rust with the buttons in the title bar.
- [Activity & Audit Trail](/sdk/activity-audit): Browse, export, and append chronological activity entries for observability and compliance.
- [Add data](/sdk/add-data): Use helper-first writes for normal MuBit memory flows, and fall back to raw ingest only when you need explicit async ingest control.
- [Agent State & Goals](/sdk/agent-state-and-goals): Manage run-scoped variables, goals, actions, and cycle history with MuBit control APIs.
- [Agent-decided Retrieval vs RAG](/sdk/associative-retrieval-vs-rag): Why MuBit lets the agent decide retrieval mode and compartment instead of forcing you to hard-code a RAG pipeline.
- [Associative Retrieval](/sdk/associative-retrieval): Use MuBit routed retrieval and context assembly to produce grounded, scoped answers and reusable prompt context.
- [Authentication (API key etc)](/sdk/authentication): Configure MuBit API-key authentication correctly for secure and stable SDK integrations.
- [Control Plane (State Runtime)](/sdk/control-plane): Use MuBit control APIs as the stable integration boundary for memory, context assembly, diagnostics, planning state, and coordination.
- [Data Guardrails (ACL related)](/sdk/data-guardrails): Protect memory access with route-level policy gates and node-level ACL governance.
- [Data Plane (Memory)](/sdk/data-plane-memory): Understand direct-core memory operations, policy boundaries, and when low-level lanes are appropriate.
- [Framework Integrations](/sdk/framework-integrations): Native MuBit memory adapters for popular agent frameworks. This page is a reference index; full walkthroughs live on the per-framework pages.
- [Full Fledged Support Agent](/sdk/full-support-agent): Build an end-to-end support agent with helper-first memory, compaction-safe context, coordination, and reinforcement.
- [gRPC Transport Guide](/sdk/grpc-transport): When to use gRPC vs HTTP, endpoint configuration, TLS, and proto file locations for MuBit SDKs.
- [How MuBit works](/sdk/how-mubit-works): How your agents go from stateless to self-improving with MuBit's memory loop.
- [LLM Provider Support](/sdk/llm-providers): Which LLM client libraries mubit.learn auto-instruments, and what each integration captures.
- [Observability, Consistency, and Operations](/sdk/observability): Operate MuBit safely with ingest/query/context telemetry, memory diagnostics, checkpoint discipline, and explicit agent coordination traces.
- [Projects, Agents, Skills, Prompts](/sdk/projects-and-agents): The Managed MuBit resource model — group agents into Projects, version their prompts and skills, and ship changes safely with a candidate → active promotion flow.
- [Pub-Sub & Semantic Subscriptions](/sdk/pubsub-and-events): Build event-driven workflows with run-scoped control streams and semantic subscriptions on the data plane.
- [Query patterns](/sdk/query-patterns): Use routed retrieval, direct semantic search, and control filters for structured MuBit retrieval.
- [Rate Limits](/sdk/rate-limits): Per-instance quotas, the headers MuBit returns on every response, and how to size your client backoff.
- [Retries and Idempotency](/sdk/retries): When to retry MuBit calls, how to make writes safe to repeat, and what the SDK does for you automatically.
- [Retrieve data](/sdk/retrieve-data): Use recall for answer-oriented retrieval and getContext for model-ready context assembly, with raw query only for explicit control.
- [SDK Configuration Reference](/sdk/sdk-configuration): Environment variables, transport selection, endpoint defaults, and connection options for MuBit SDKs.
- [SDK Helpers](/sdk/sdk-helpers): Explicit helper methods for memory, context assembly, and the learning loop — when you outgrow the mubit.learn drop-in.
- [SDK methods](/sdk/sdk-methods): Three integration layers — from zero-config closed-loop to full low-level control.
- [Sessions & Branching](/sdk/sessions-and-branching): Use checkpoints for durable task continuity and core sessions for reversible scratch branches.
- [Troubleshooting](/sdk/troubleshooting): Common issues, error messages, and solutions for MuBit SDK and API usage.
- [Examples / Cookbook](/sdk/tutorials): Start with the complete multi-SDK implementation first, then use cookbooks for focused patterns.
- [Webhooks](/sdk/webhooks): Receive events from MuBit when ingest jobs complete, lessons are extracted, or handoffs change state.
- [Agno Integration](/sdk/integrations/agno): Use MuBit as a persistent memory backend and toolkit for Agno agents. MemoryDb plus an LLM-callable Toolkit.
- [CrewAI Integration](/sdk/integrations/crewai): Route CrewAI's unified Memory system through MuBit. Agent observations persist across runs and crews.
- [Google ADK Integration](/sdk/integrations/google-adk): Plug MuBit into ADK's Runner as a BaseMemoryService. Session events auto-ingest; memory search enriches agent context.
- [LangChain Integration](/sdk/integrations/langchain): Drop-in BaseMemory for any LangChain chain. Cross-session memory works automatically through MuBit's semantic retrieval.
- [LangGraph Integration](/sdk/integrations/langgraph): Use MuBit as a persistent BaseStore for LangGraph StateGraphs. Each node reads or writes through PutOp / SearchOp.
- [LlamaIndex Integration](/sdk/integrations/llama): Use MuBit as a chat-memory store and a vector store for LlamaIndex indexes. Two surfaces, one memory pool.
- [MCP Server](/sdk/integrations/mcp): Expose MuBit as 15 tools over the Model Context Protocol stdio transport. Usable from Claude Desktop, Cursor, and any MCP client.
- [Vercel AI SDK Integration](/sdk/integrations/vercel-ai-sdk): Middleware that wraps any AI SDK model with automatic memory injection and interaction capture.
- [Branching Memory (What-If Simulations)](/recipes/branching-memory): Use checkpoints and core session branches for reversible what-if exploration without polluting the main execution path.
- [Building Event-Driven Agents](/recipes/event-driven-agents): Use control events to react to memory changes, checkpoints, outcomes, and coordination artifacts instead of relying only on polling loops.
- [Lane-Scoped Multi-Agent Memory](/recipes/lane-scoped-memory): Isolate agent memory into named lanes so specialists see only their relevant context while sharing a common run.
- [Multi-Agent Shared State](/recipes/multi-agent-shared-state): Coordinate specialist agents with agent registration, scoped access, handoffs, feedback, and shared context assembly.
- [Prompt Optimization Lifecycle](/recipes/prompt-optimization): Capture outcomes → optimize → diff → activate. A human-in-the-loop workflow for evolving agent prompts from real execution data.
- [Stateful Task Trees](/recipes/stateful-task-trees): Build planner task trees with goals, actions, cycles, and checkpoints so orchestration survives retries and compaction.
- [Step-Level Outcomes and Process Rewards](/recipes/step-level-outcomes): Record per-step reward signals during agentic runs so reflection produces richer, step-attributed lessons.
- [Support Agent Memory Loop](/recipes/support-agent-loop): Build a support-agent loop with context assembly, checkpointing, reflection, and outcome reinforcement.
- [Temporal Memory Patterns](/recipes/temporal-memory-patterns): Use occurrence time, temporal range queries, staleness detection, and budget control for time-aware agent memory.
- [Control gRPC API](/api-reference/control-grpc): Current public MuBit ControlService RPC surface for memory, diagnostics, learning loop, coordination, and explicit planning state.
- [Control HTTP API](/api-reference/control-http): Current public MuBit control-plane HTTP routes for memory, context, diagnostics, coordination, and explicit planning state.
- [Core Direct Lanes and Policy](/api-reference/core-direct-lanes): Policy contract for direct core query routes and safe rollout guidance.
- [Errors](/api-reference/errors): HTTP status codes, error types, and structured error payloads returned by the MuBit control API.
- [State Management Surface](/api-reference/state-management): Run-scoped state routes for variables and concepts, plus deprecated planning state (goals / actions / cycles).