OSSP-AGI — Protocol Discovery Workbench
A discovery sprint in protocol thinking: OSSP-AGI is an 18-prototype experiment that explored what an agent-native protocol ecosystem would require—spanning API, data, event, semantic, and agent-to-agent layers. The result is a reference architecture, launchable workbench, and roadmap toward the MCP Hub.
TL;DR
- Hypothesis: Could a single developer prototype the full protocol stack needed for autonomous agent collaboration—including data, events, semantics, and MCP orchestration?
- Experiment: Built 18 interconnected prototypes that stress-tested API schemas, event routing, semantic manifests, and agent handshakes inside a local-first discovery lab.
- Impact: Identified the core protocol set that now powers the MCP Hub roadmap, produced reusable widgets, and documented the architectural blueprint for agent-native systems.
Hypothesis & Experiment Design
OSSP-AGI started as a thought experiment: abandon the monolithic "all-in-one app" and instead model the protocol lattice autonomous agents would actually need. The application became a lab for breadth-first exploration.
| Protocol Layer | Focus | Prototype Highlights |
|---|---|---|
| API | Contract discovery, manifest normalization | Spec ingestion, diffing, and URN-addressable registries |
| Data | Schema lineage, typed payload guarantees | Columnar telemetry, semantic versioning, provenance trails |
| Event | Cross-protocol signaling, replay safety | Temporal routing, failure semantics, and idempotent schedulers |
| Semantic | Context vocabularies, intent encoding | Mission schemas, capability taxonomies, translation matrices |
| Agent / MCP | Agent-to-agent negotiation, tooling surfaces | Conversation broker, MCP tool skeletons, policy enforcement |
Each prototype fed the next, revealing how protocols needed to compose rather than compete. The exercise validated that ecosystem architecture—not a single product—was the real deliverable.
Key Discoveries
- Ambition gravity: Attempting an all-in-one product created focus drift; treating the workbench as a protocol lab surfaced the reusable core faster.
- MCP nucleus: The API, data, event, semantic, and A2A layers formed the durable center—everything else was negotiable.
- Composable wins: Pairing semantic manifests with event routing unlocked dependable orchestration experiments.
- Hard lessons: Persisting agent state across MCP sessions exposed serialization pitfalls that now inform the Hub design.
“Stupid as a product, invaluable as a thought experiment.” OSSP-AGI proved that the protocols deserve to stand alone.
Launch Bundle & Evidence
Even for an experiment, credibility mattered. The v0.25 workbench ships with reproducible artifacts so external reviewers can replay the discovery journey.
- Automation:
npm run demo:preflightregenerates manifests, telemetry, and walkthrough scripts from scratch. - Artifacts:
artifacts/launch/v0.25/packages protocol diagrams, dependency graphs, and validation transcripts. - Security posture: Hardened IAM policies, fail-closed runtime checks, and deterministic error responses keep the lab honest.
- External specs: OpenAPI/AsyncAPI scrapes are versioned and traceable via scripted pulls.
From Experiment to Ecosystem
OSSP-AGI now anchors the roadmap for the next wave of tooling. The lessons became action items.
- MCP Hub: Porting the validated core protocols into a dedicated Model Context Protocol server.
- Open-source widgets: Extracting the most useful prototypes—manifest diffing, schema telemetry, mission splitters—into standalone tools.
- Reference architecture: The case study doubles as documentation for teams exploring agent-native ecosystems.
What’s Next
- Focus the catalog: Sunset breadth experiments and deepen the validated protocol set.
- Operationalize MCP tooling: Bring the Agent/MCP prototypes to production readiness with policy governance.
- Community releases: Package the discovery scripts and telemetry viewers as public learning resources.