Framework Support

This page is generated from the versioned files in frameworks/. It records the declared support contract; runtime availability is still reported by agentbridge coverage-report.

Summary

Framework status	Count
`blocked`	1
`partial`	6
`verified`	2

Integrations

Framework	Package	Status	Verification lanes	Declared capabilities
Amazon Bedrock AgentCore	`agentbridge-agentcore` `>=0.1,<1`	`partial`	`offline`, `optional_dependency`, `credentialed`	9
CrewAI	`crewai` `>=0.11.2,<0.12`	`blocked`	`offline`, `optional_dependency`, `credentialed`	8
LangChain Deep Agents	`deepagents` `>=0.7,<1`	`partial`	`offline`, `optional_dependency`, `credentialed`	9
Google ADK	`google-adk` `>=2.9,<3`	`partial`	`offline`, `optional_dependency`, `credentialed`	10
LangChain	`langchain` `>=1.4,<2`	`partial`	`offline`, `optional_dependency`, `credentialed`	13
LangGraph	`langgraph` `>=1.2.11,<2`	`verified`	`offline`, `optional_dependency`, `credentialed`	13
OpenAI Agents SDK	`openai-agents` `>=0.20,<0.21`	`partial`	`offline`, `optional_dependency`, `credentialed`	10
Pydantic AI	`pydantic-ai-slim` `>=2.42,<3`	`verified`	`offline`, `optional_dependency`, `credentialed`	11
Strands Agents	`strands-agents` `>=1.55,<2`	`partial`	`offline`, `optional_dependency`, `credentialed`	11

Amazon Bedrock AgentCore

Framework ID: agentcore
Plugin: plugins/agentbridge-agentcore
Next slice: Add IAM-aware runtime, Memory, Gateway, and deployment smoke tests.
Native escape hatches: runtime client, memory, gateway, identity, native AWS operations
Upstream review sources: AgentCore documentation

Capability	Declared support
`agent.model`	`extension`
`agui.events`	`extension`
`deployment.serverless`	`extension`
`observability.diagnostics`	`extension`
`observability.tracing`	`extension`
`state.memory`	`extension`
`state.session`	`extension`
`streaming.events`	`extension`
`tools.mcp`	`extension`

CrewAI

Framework ID: crewai
Plugin: plugins/agentbridge-crewai
Next slice: Verify the existing plugin in an isolated Python and dependency environment.
Native escape hatches: crew, task, process, native agent options
Upstream review sources: CrewAI documentation

Capability	Declared support
`agent.instructions`	`extension`
`agent.model`	`extension`
`human_approval`	`extension`
`observability.raw`	`extension`
`state.memory`	`extension`
`tools.sync`	`extension`
`workflow.delegation`	`extension`
`workflow.roles_tasks`	`extension`

LangChain Deep Agents

Framework ID: deepagents
Plugin: plugins/agentbridge-deepagents
Next slice: Add credentialed sandbox and hosted deployment execution lanes.
Native escape hatches: sandbox, filesystem backend, subagents, skills
Upstream review sources: Deep Agents documentation

Capability	Declared support
`agent.instructions`	`full`
`agent.model`	`extension`
`deployment.serverless`	`native_only`
`human_approval`	`extension`
`state.checkpointing`	`extension`
`state.memory`	`extension`
`structured_output`	`extension`
`tools.sync`	`extension`
`workflow.handoffs`	`extension`

Google ADK

Framework ID: google_adk
Plugin: plugins/agentbridge-google-adk
Next slice: Run credentialed eval and deployment publishing conformance in an isolated telemetry environment.
Native escape hatches: Agent, Runner, session service, artifact service, deployment metadata
Upstream review sources: Google ADK documentation

Capability	Declared support
`agent.instructions`	`full`
`agent.model`	`extension`
`deployment.serverless`	`extension`
`evals`	`extension`
`state.memory`	`extension`
`state.session`	`extension`
`streaming.events`	`extension`
`structured_output`	`extension`
`tools.sync`	`full`
`workflow.handoffs`	`extension`

LangChain

Framework ID: langchain
Plugin: plugins/agentbridge-langchain
Next slice: Graduate high-value hosted deployment, provider, and observability paths with credentialed tests.
Native escape hatches: native_options, native model, middleware, callbacks
Upstream review sources: LangChain documentation, LangChain docs index

Capability	Declared support
`agent.instructions`	`full`
`agent.model`	`extension`
`evals`	`extension`
`human_approval`	`extension`
`observability.tracing`	`extension`
`state.memory`	`extension`
`state.session`	`extension`
`streaming.events`	`extension`
`structured_output`	`extension`
`tools.async`	`extension`
`tools.mcp`	`extension`
`tools.sync`	`full`
`workflow.graph`	`native_only`

LangGraph

Framework ID: langgraph
Plugin: agentbridge.adapters.langgraph
Next slice: Credentialed remote graph, deployment, and hosted runtime conformance.
Native escape hatches: native_graph, native_options, runtime_config
Upstream review sources: LangGraph documentation

Capability	Declared support
`agent.instructions`	`partial`
`agent.model`	`extension`
`human_approval`	`extension`
`observability.diagnostics`	`full`
`observability.tracing`	`extension`
`state.checkpointing`	`extension`
`state.session`	`extension`
`streaming.events`	`full`
`structured_output`	`full`
`tools.async`	`full`
`tools.sync`	`full`
`workflow.graph`	`full`
`workflow.routing`	`extension`

OpenAI Agents SDK

Framework ID: openai_agents
Plugin: plugins/agentbridge-openai-agents
Next slice: Resolve the OpenAI dependency-major conflict and verify native resume execution.
Native escape hatches: Agent, Runner, handoffs, guardrails, MCP servers
Upstream review sources: OpenAI Agents SDK documentation

Capability	Declared support
`agent.instructions`	`full`
`agent.model`	`extension`
`guardrails`	`extension`
`human_approval`	`extension`
`observability.tracing`	`extension`
`streaming.events`	`partial`
`structured_output`	`extension`
`tools.mcp`	`extension`
`tools.sync`	`full`
`workflow.handoffs`	`extension`

Pydantic AI

Framework ID: pydantic_ai
Plugin: agentbridge.adapters.pydantic_ai
Next slice: Expand durable execution, provider smoke tests, and native evaluation integration.
Native escape hatches: agent_options, model_settings, deps
Upstream review sources: Pydantic AI documentation

Capability	Declared support
`agent.instructions`	`full`
`agent.model`	`full`
`observability.diagnostics`	`full`
`observability.raw`	`full`
`runtime.retries`	`extension`
`state.session`	`partial`
`streaming.events`	`full`
`structured_output`	`full`
`structured_output.validation_retries`	`extension`
`tools.async`	`extension`
`tools.sync`	`full`

Strands Agents

Framework ID: strands
Plugin: plugins/agentbridge-strands
Next slice: Add credentialed AWS and AgentCore deployment conformance.
Native escape hatches: Agent options, MCP client, hooks, interventions, AgentCore metadata
Upstream review sources: Strands documentation

Capability	Declared support
`agent.instructions`	`full`
`agent.model`	`extension`
`deployment.serverless`	`extension`
`guardrails`	`extension`
`observability.tracing`	`extension`
`state.memory`	`extension`
`state.session`	`extension`
`streaming.events`	`extension`
`structured_output`	`extension`
`tools.mcp`	`extension`
`tools.sync`	`full`