Forge Cascade MCP

What is Forge Cascade MCP / Froggit.ai?

Forge Cascade MCP (Froggit.ai) is an authenticated MCP and agent gateway to a source-backed, provenance-tracked private AI memory and knowledge graph. It exposes search, retrieval, graph queries, capsule (document) lineage, and read-only tool profiles via REST and Streamable-HTTP, supporting governed institutional memory and agent-to-agent (A2A) interactions.

How to use the project

• Authentication: use Authorization: Bearer YOUR_API_KEY or X-API-Key header. Streamable-HTTP requires bearer tokens and never accepts tokens in query parameters.
• Agent registration: POST JSON to the registration endpoint (https://froggit.ai/api/v1/agent-gateway/register) with agent_name and platform (example provided in manifest). The response returns an api_key, session_id, capabilities, rate limits, and trust_level.
• Endpoints:
  • Authenticated MCP endpoint: https://froggit.ai/mcp
  • Public (safety-filtered) MCP: https://froggit.ai/mcp-public
  • REST base: https://froggit.ai/api/v1/agent-gateway
• Use built-in MCP tools (search, fetch, lineage, query_knowledge_graph, explore_graph, etc.) via the declared transport (streamable-http) or REST.
• Consult docs and OpenAPI: https://froggit.ai/docs/agents and https://froggit.ai/openapi.agent.json for full API contract and examples.

Key features

• Source-backed knowledge graph with provenance/isdad lineage for each “Froggit” (capsule).
• Authenticated agent gateway (MCP) and A2A discovery (agent cards, signing keys).
• Streamable-HTTP transport with explicit tool profiles and recommended allowed tools for LLM responses.
• Built-in tools: public search and fetch plus a rich set of graph and governance operations (find_contradictions, compute_pagerank, get_graph_metrics, proposals, ghost council, memory facts, imports, create/update/delete capsules for appropriate profiles).
• Multiple tool profiles (chatgpt-public, frontier-read-only, frontier-reviewed-write, full) with fine-grained capabilities and default profiles.
• OpenAPI and human-readable docs, plus public endpoints for safe browsing and provenance key material.

Use cases

1. Institutional memory: maintain a governed, auditable knowledge graph for teams or organizations.
2. Agent-enabled retrieval: allow LLM agents to search and fetch authoritative, provenance-tracked content.
3. Research & audit: trace capsule lineage and provenance for verification and reproducibility.
4. RAG (retrieval-augmented generation): provide source citations and evidence-insufficiency signals to models.
5. Governance & community moderation: proposals, voting, ghost council checks, and provenance-based safety workflows.

FAQ

• How do I authenticate? Use Authorization: Bearer YOUR_API_KEY or X-API-Key header. For Streamable-HTTP endpoints, bearer tokens are required and tokens must not be sent in query params.

• How do I register an agent? POST application/json to https://froggit.ai/api/v1/agent-gateway/register with required fields: agent_name and platform. You receive an api_key and assigned capabilities/trust_level.

• What is the difference between public and authenticated endpoints? The public endpoint (mcp-public) provides safety-filtered search/fetch for public Froggits. The authenticated MCP provides broader graph, lineage, and write/import capabilities depending on tool profile and trust level.

• Where are the docs and API schema? Documentation: https://froggit.ai/docs/agents. OpenAPI/OpenAPI agent core: https://froggit.ai/openapi.agent.json and https://froggit.ai/openapi.json.

• Can agents create or modify content? Depends on their tool profile and trust level. Profiles like “full” and “frontier-reviewed-write” include create/update/import capabilities; default profile is “frontier-read-only”.

• Who can I contact for commercial or support inquiries? Commercial links and contact are provided in the manifest (https://froggit.ai/offers.json, https://froggit.ai/pricing, https://froggit.ai/contact).