Features¶
Lango provides a comprehensive set of features for building intelligent AI agents. This section covers each subsystem in detail.
-
Multi-provider support for OpenAI, Anthropic, Gemini, and Ollama with a unified interface and automatic fallback.
-
Connect your agent to Telegram, Discord, and Slack. Manage conversations across channels from a single instance.
-
Self-learning knowledge store with 8-layer context retrieval, pattern recognition, and agent learning tools.
-
Automatic conversation compression through observations and reflections for long-running sessions.
-
Vector embeddings with OpenAI, Google, or local providers. Retrieval-augmented generation for semantic context injection.
-
BoltDB-backed triple store with hybrid vector + graph retrieval for deep contextual understanding.
-
Typed knowledge ontology with schema lifecycle governance, temporal truth maintenance, entity resolution, ACL, and P2P exchange.
-
Hierarchical sub-agents (Operator, Navigator, Vault, Librarian, Automator, Planner, Chronicler) working together on complex tasks.
-
Agent-to-Agent protocol for remote agent discovery and inter-agent communication.
-
Decentralized agent-to-agent connectivity via libp2p with DID identity, knowledge firewall, and ZK-enhanced handshake.
-
Budget management, risk assessment, dynamic pricing, P2P negotiation, and milestone-based escrow for agent commerce.
-
EVM smart contract interaction with ABI caching, view/pure reads, and state-changing calls.
-
ERC-7579 modular smart accounts with session keys, ERC-4337 paymaster support, and on-chain policy enforcement.
-
Token usage tracking, health monitoring, audit logging, and metrics endpoints for operational visibility.
-
Per-agent persistent memory for cross-session context retention and experience accumulation.
-
File-based skills with import from URLs and GitHub repositories. Extend agent capabilities without code changes.
-
Autonomous knowledge agent that observes conversations and proactively curates the knowledge base.
-
Customizable prompt sections for agent personality, safety rules, and behavior tuning.
-
Collaborative environments where multiple agents share code, messages, and context with git bundle exchange and contribution tracking.
-
Task-scoped multi-agent collaboration with role assignment, conflict resolution, budget tracking, and payment coordination.
-
Threshold-based operational alerting with policy block rate, recovery retry, and circuit breaker monitors.
-
Policy-based command safety evaluation with shell wrapper unwrapping, opaque pattern detection, and catastrophic pattern blocking.
-
Multi-panel terminal dashboard with sidebar navigation, live metrics, and page-based UI for chat, settings, tools, and status.
-
Process isolation via macOS Seatbelt (Linux: planned, not yet enforced) for tool execution with network deny and workspace-scoped access.
-
Output sanitization that strips thought tags, internal markers, raw JSON, and custom patterns from agent responses.
-
Token-budget-aware context allocation with retrieval coordinator, config profiles (off/lite/balanced/full), and relevance auto-adjustment.
-
Pre-built configuration templates for common deployment scenarios. Quick-start your agent with sensible defaults.
-
Connect to external MCP servers for stdio, HTTP, and SSE transports. Extend agent tooling with the Model Context Protocol.
-
Durable execution engine for multi-agent task orchestration with append-only journal, PEV verification, and typed validators.
-
Persistent checkpoints, session lineage, git-aware attribution, and signed provenance bundle export/import for auditable multi-agent work.
Feature Reference Catalog¶
- AI Providers
- Channels
- Knowledge System
- Agent Response Format
- Learning Engine
- Observational Memory
- Embedding & RAG
- Knowledge Graph
- Knowledge Ontology
- Multi-Agent Orchestration
- A2A Protocol
- P2P Network
- P2P Economy
- Smart Contracts
- Smart Accounts
- Observability
- Skill System
- Proactive Librarian
- System Prompts
- Operational Alerting
- Exec Safety
- Cockpit TUI
- Config Presets
- MCP Integration
- RunLedger (Task OS)
- Session Provenance
- Zero-Knowledge Proofs
Feature Status¶
| Feature | Status | Config Key |
|---|---|---|
| AI Providers | Stable | agent.provider |
| Channels | Stable | channels.* |
| Knowledge System | Stable | knowledge.enabled |
| Observational Memory | Stable | observationalMemory.enabled |
| Embedding & RAG | Stable | embedding.* |
| Knowledge Graph | Experimental | graph.enabled |
| Knowledge Ontology | Experimental | ontology.enabled |
| Multi-Agent Orchestration | Experimental | agent.multiAgent |
| A2A Protocol | Experimental | a2a.enabled |
| P2P Network | Experimental | p2p.enabled |
| P2P Economy | Experimental | economy.enabled |
| Smart Contracts | Experimental | payment.enabled |
| Smart Accounts | Experimental | smartAccount.enabled |
| Observability | Experimental | observability.enabled |
| Skill System | Stable | skill.enabled |
| Proactive Librarian | Experimental | librarian.enabled |
| System Prompts | Stable | agent.promptsDir |
| Agent Memory | Experimental | agentMemory.enabled |
| P2P Workspaces | Experimental | p2p.workspace.enabled |
| P2P Teams | Experimental | p2p.enabled + team coordination |
| Config Presets | Stable | lango onboard --preset |
| MCP Integration | Stable | mcp.enabled |
| RunLedger (Task OS) | Experimental | runLedger.enabled |
| Session Provenance | Experimental | provenance.enabled |
| Operational Alerting | Experimental | alerting.enabled |
| Exec Safety | Stable | hooks.blockedCommands |
| Cockpit TUI | Stable | — |
| OS-level Sandbox | Experimental | sandbox.enabled |
| Response Gatekeeper | Stable | gatekeeper.enabled |
| Context Engineering | Stable | context.*, retrieval.*, contextProfile |
| Tool Hooks | Experimental | hooks.enabled |
| Tool Catalog | Internal | — |
| Event Bus | Internal | — |
Experimental Features
Features marked as Experimental are under active development. Their APIs, configuration keys, and behavior may change between releases. Enable them explicitly via their config flags.