AgentOS MCP Bridge

What It Is

The architectural connector between AgentOS's internal scope structure and external AI hosts. AgentOS uses 13 scoped MCP servers (VC1-7 for value chain, SA1-6 for support activities) for internal data boundaries, but MCP requires a single endpoint per deployment. The bridge multiplexes across value-chain-scoped handlers internally while presenting one server to external hosts. The UI communicates through JSON-RPC messages proxied by the bridge, with the React frontend built to a single-file HTML bundle via vite-plugin-singlefile for MCP App iframe embedding.

Why It Matters

Internal scope boundaries and external interface simplicity pull in opposite directions. Resolving this tension is the prerequisite for commercial distribution.

MCP Apps, Claude, ChatGPT, VS Code -- each connects to one URL. Thirteen separate servers means thirteen separate integrations per host, which defeats the "one build, every surface" principle. The MCP Bridge solves this by running a single MCP server process that internally multiplexes across scoped handlers. The scope resolver runs inside the server, not between servers. External hosts see one endpoint. Internal code sees scoped data boundaries. Without this bridge, AgentOS's governance capabilities are locked to controlled environments. With it, constitutional governance becomes a plugin on any MCP-compatible host.

Proof Points

MCP Architecture Spec v2.0: authoritative specification, supersedes 3 prior specifications that had conflicting topologies
Three breaking changes resolved in spec: server topology (13 servers to 1), UI shell to MCP App embedding, repository structure consolidation
Single-file HTML bundle via vite-plugin-singlefile for MCP App iframe embedding
JSON-RPC communication protocol through bridge -- not direct HTTP calls
Scope resolver adapted from SovereignAI codebase for unified server architecture
Status: 0% implemented -- the single biggest architectural gap in AgentOS (~52% overall completion)
Specified as the critical path for AgentOS commercial distribution

Market Position and IP

The MCP Bridge is the critical missing piece for AgentOS distribution. No competing framework has solved the scope-multiplexing problem for MCP -- presenting multiple internal scope boundaries as a single external endpoint while maintaining data isolation. First-mover advantage applies: as MCP becomes the standard protocol for AI tool integration, the bridge pattern will be required by any multi-scope agent system. The bridge enables constitutional governance as a deployable plugin on Claude, ChatGPT, VS Code, and any future MCP-compatible host -- converting AgentOS from a standalone system to a distributable product.

Novel Research Contribution

Existing MCP research treats each server as an independent tool provider. The MCP Bridge demonstrates that scope multiplexing -- running multiple logically separated tool sets through a single protocol endpoint -- is architecturally necessary for enterprise agent systems with data boundaries. The pattern resolves the tension between MCP's single-endpoint-per-host constraint and enterprise requirements for scoped data access. This is a novel contribution to the MCP protocol ecosystem.

Implementation and Impact

When built, the MCP Bridge will enable AgentOS deployment on any MCP-compatible surface without per-host integration work. The deliverable: a single MCP server endpoint that exposes all 13 scoped tool sets with the Governance Engine UI embeddable as an MCP App. This is the architectural prerequisite for commercial distribution of AgentOS -- the difference between a research system and a product. Measurable outcome: AgentOS governance available as a single-click install on Claude, ChatGPT, VS Code, and custom MCP hosts.

Connections

Imperatives: Fractal Design, Constraint Surface Governance
Builds: AgentOS (core dependency), Self-Learning Agent (integration path from Python prototype)
Papers: brick-contract
Capabilities: Agentic System of Systems