- Rust workspace with 12 crates
-
agentos-core- Fundamental types, error types, capability enums -
agentos-identity- Agent identity with capability-based permissions -
agentos-memory- Persistent memory layer with keyword search, filtering -
agentos-artifacts- Artifact storage for any binary data -
agentos-messaging- Async agent messaging with priority, inbox, read tracking -
agentos-tasks- Task lifecycle: Pending -> InProgress -> Completed/Failed/Cancelled -
agentos-knowledge- Knowledge graph: entity nodes + typed relations -
agentos-workspace- Top-level orchestrator unifying all layers with capability-gated operations -
agentos-sdk- Facade crate re-exporting all modules viaagentos_sdk::prelude::*
-
agentos-walrus- WalrusConfig (testnet presets), WalrusClientWrapper - WalrusMemoryStore - MemoryStore implementation persisting to Walrus blobs
- WalrusArtifactStore - ArtifactStore implementation storing binary data on Walrus
-
agentos-persistence- WorkspaceSnapshot (full state serialization) - PersistenceManager - snapshot/restore/load workspace from Walrus
- Snapshot roundtrip tests (serialize -> deserialize -> restore)
-
agentos-cli- CLI with subcommands: init, register-agent, remember, search, snapshot, status, create-task
-
agentos-api- axum-based REST API server with 20+ endpoints - All CRUD endpoints for workspaces, agents, memories, artifacts, messages, tasks, knowledge
- Snapshot/restore endpoints
- CORS enabled for cross-origin access
- Server starts on 0.0.0.0:3000
- 7 integration tests passing (5 StudyTogether workflow + 2 persistence)
- All crates compile with zero warnings
- Framework adapters (LangGraph, CrewAI, OpenAI Agents SDK)
- Seal integration for encrypted agent knowledge
- Sui Move contracts for on-chain agent identity
- MemWal memory indexing integration
- Docker Compose setup for local development
- Python SDK / bindings