Console adapters are thin wrappers around the canonical @orangecheck/agent-core package. Adding one to your stack is a single import and a one-line server-side verifier middleware. Every adapter is open-source MIT under github.com/orangecheck.
Model Context Protocol
A scope-aware MCP wrapper. Every tool call carries the active OC Agent delegation; every result emits an agent-action envelope (OC Stamp shape) anchored to Bitcoin via the OC Stamp pipeline.
v1 ships an `@orangecheck/agent-mcp` adapter and a runnable demo. MCP is the highest-leverage layer because the rest of the stack is converging on it.
Anthropic Tool Use
Wrap Claude tool definitions with OC scope grammar. Tool execution is gated by the live delegation state; receipts feed the audit bundle.
OpenAI Responses / Function Calling
Function-calling adapter that emits agent-action envelopes per invocation. Compatible with both Responses and the older Chat Completions tool path.
Vercel AI SDK
Drop-in `tool()` wrappers and a server-side verifier middleware. Intended to be a one-import switch for teams already on the AI SDK.
LangGraph
Node-level scope enforcement and a graph-replay tool that re-derives every action receipt from the persisted graph state.
Salesforce Agentforce
Enterprise-grade audit bundles for Agentforce-driven workflows, mappable to the Salesforce data model for inline compliance review.
Microsoft Copilot Studio
Connector + audit-trail integration for Copilot Studio agents. Pulled forward if a design partner names it.
Adapter prioritization is design-partner-driven. If your team is on a stack we haven't listed, tell us — “planned” integrations move to “shipping” when a design partner needs them.