All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
0.7.0 - 2026-03-21
- Claude Agent SDK adapter (
agentanycast[claude]) — expose prompt-based agents over P2P - AWS Strands Agents adapter (
agentanycast[strands]) — bridge Strands agents to P2P - Proto stubs updated with OpenTelemetry trace context fields for distributed tracing
0.6.0 - 2026-03-20
- MCP server mode with stdio and HTTP transports
- AGNTCY Directory client for agent discovery via the AGNTCY ecosystem
- Streaming support for chunked artifact delivery
did:webanddid:dnsDID method support alongside existingdid:keyBaseAdapterABC for consistent adapter authoring with auto-card generation
- Adapter robustness improvements and MCP server correctness fixes
0.5.0 - 2026-03-19
- Google ADK adapter (
agentanycast[google-adk]) - OpenAI Agents SDK adapter (
agentanycast[openai-agents]) - A2A v1.0 protocol compatibility layer (
task_to_a2a_json/task_from_a2a_json) - OASF record conversion (
card_to_oasf_record/card_from_oasf_record) - AI discoverability metadata for package indexes
0.4.0 - 2026-03-19
- CrewAI adapter (
agentanycast[crewai]) — serve CrewAI crews as P2P agents - LangGraph adapter (
agentanycast[langgraph]) — serve compiled graphs as P2P agents - Click-based CLI with
demo,discover,send,status, andinfocommands - Integration test framework with Docker Compose E2E tests
0.3.0 - 2026-03-18
- HTTP Bridge support —
url=parameter insend_task()for interop with standard HTTP A2A agents - MCP tool mapping (
mcp_tools_to_agent_card,mcp_tool_to_skill,skill_to_mcp_tool) - DID support (
peer_id_to_did_key,did_key_to_peer_id) - PyPI release workflow with OIDC Trusted Publishing
0.2.0 - 2026-03-18
- Skill-based anycast discovery via
discover()method - Relay connection support for cross-network communication
connect_peer()andlist_peers()for peer management- Real P2P task E2E tests over Noise-encrypted channels
0.1.0 - 2026-03-17
Nodeclass — async context manager for P2P agent communicationAgentCardandSkilldataclasses for agent capability descriptionTaskHandlefor tracking outgoing tasks withwait()supportIncomingTaskfor receiving and responding to tasks (complete(),fail(),update_status())- Automatic daemon lifecycle management (start, stop, health check)
- Auto-download of daemon binary when not provided
- gRPC client for SDK-daemon communication over Unix domain socket
homeparameter for running multiple nodes on the same machine- Hello world example with server and client agents
- PEP 561
py.typedmarker for type checking support - CI pipeline with ruff lint, mypy type check, and multi-version pytest