prism-mcp
Provides a tool to import Discord conversations and content into Prism projects, enabling agents to bring external Discord data into the knowledge base.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@prism-mcpgenerate an image of a serene mountain lake at sunset"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
@skyphusion/prism-mcp
License: MIT
Version: 1.0.0
API: prism (AGPL playground Worker)
Control plane (metered commercial door): prism-control-plane
Native clients: prism-ios, prism-android
Agent MCP for Prism: HTTP parity with the prism /api/*
surface so coding agents can chat, generate images/video/music/TTS, manage RAG documents and
projects, compact long threads, and poll long jobs -- without a browser.
Stateless Model Context Protocol Worker that proxies curated tools to a prism instance over HTTPS.
Parity: docs/PARITY.md -- full HTTP agent parity in 1.0.0; live Flux mic WebSocket is the only major human SPA gap.
How the pieces fit together
flowchart TB
subgraph agents["MCP clients"]
CC["Claude Code / Cursor / any MCP client"]
end
subgraph mcp["Worker: prism-mcp<br/>prism-mcp.skyphusion.org"]
Gate["Bearer MCP_TOKEN"]
Tools["32 curated tools + read/write escape hatch"]
Proxy["HTTPS proxy<br/>Cookie __Host-prism_session"]
Gate --> Tools --> Proxy
end
subgraph play["Worker: prism / skyphusion-llm<br/>play.skyphusion.org"]
Auth["Session auth"]
API["/api/* HTTP surface"]
RAG["RAG · projects · compact"]
Jobs["LongRunWorkflow<br/>video · music"]
STTws["WS /api/stt/stream<br/>Flux live voice"]
Auth --> API
API --> RAG
API --> Jobs
end
subgraph plane["Optional hybrid spend"]
PCP["prism-control-plane<br/>play-proxy · pcp_ in prefs"]
end
subgraph cf["Cloudflare AI"]
GW["AI Gateway"]
WAI["Workers AI + Unified Billing"]
GW --> WAI
end
CC -->|"tools/call<br/>Streamable HTTP"| Gate
Proxy -->|"PRISM_URL + session cookie"| Auth
API --> GW
Jobs --> GW
Auth -.->|"prefs may hold pcp_"| PCP
PCP -.-> GW
STTws -.->|"not MCP-transportable<br/>v1.0 gap"| agentsAuth and credential split
sequenceDiagram
participant A as Agent
participant M as prism-mcp
participant P as prism
Note over A,M: Agent never sees the prism cookie
A->>M: Authorization: Bearer MCP_TOKEN
M->>M: gate token
M->>P: Cookie __Host-prism_session=PRISM_SESSION
P-->>M: /api/* JSON / SSE / binary
M-->>A: MCP tool resultTypical tool call (chat)
sequenceDiagram
participant A as Agent
participant M as prism-mcp
participant P as prism
participant G as AI Gateway
A->>M: tools/call chat {model, messages, …}
M->>P: POST /api/chat (+ session cookie)
P->>G: env.AI.run / provider
G-->>P: completion / artifact meta
P-->>M: JSON body
M-->>A: text content blockLong-run job (video / music)
sequenceDiagram
participant A as Agent
participant M as prism-mcp
participant P as prism
participant W as LongRunWorkflow
A->>M: tools/call chat (video/music model)
M->>P: POST /api/chat
P->>W: enqueue
P-->>M: job id (or pending)
M-->>A: job id
loop until done
A->>M: tools/call poll_job {id}
M->>P: GET /api/job/:id
P-->>M: status / result
M-->>A: status
end
opt artifact
A->>M: tools/call get_artifact {path}
M->>P: GET /api/artifact/*
P-->>A: bytes or URL meta
endCompact multi-turn context
flowchart LR
L["list / get_conversation"] --> C["compact_conversation"]
C --> Chat["chat with conversation_id"]
Chat --> Clear["clear_conversation_compact"]
Clear --> ChatStack position (vs native / control plane)
flowchart LR
subgraph human["Human paths"]
SPA["Browser SPA<br/>play.skyphusion.org"]
iOS["prism-ios / android<br/>Bearer pcp_"]
end
subgraph agent["Agent path"]
MCP["prism-mcp<br/>Bearer MCP_TOKEN"]
end
subgraph backends["Backends"]
Prism["prism<br/>history · RAG · compact"]
Plane["control plane<br/>meter · enroll · store"]
end
SPA --> Prism
MCP --> Prism
iOS --> Plane
Prism -.->|"optional pcp_ prefs"| PlaneRelated MCP server: litellm-mcp
Install
npm install @skyphusion/prism-mcpDeploy (Cloudflare Workers)
main = "node_modules/@skyphusion/prism-mcp/dist/mcp.js"See docs/mcp.md for secrets, session seeding, agent wiring, and the tool catalog.
Binding | Kind | Role |
| var | Prism base URL (e.g. |
| secret | Raw |
| secret | Agent gate ( |
Optional: PRISM_ACCESS_EMAIL, PRISM_ACCESS_CLIENT_ID, PRISM_ACCESS_CLIENT_SECRET for
Access-mode / Access-fronted self-hosts.
Tools (1.0.0)
Area | Tools |
Health |
|
Catalog / prefs |
|
Generation |
|
History |
|
Conversations |
|
RAG |
|
Projects | CRUD, attach/detach docs, |
Jobs / media |
|
Escape |
|
32 curated tools + prism_request_read / prism_request_write. Full route map: docs/mcp.md.
Not a tool: WebSocket /api/stt/stream (live Flux). Approximate with STT chat + tts.
Package layout
Import | Role |
| Default Worker export ( |
|
|
| Tool catalog + |
Docs
Doc | Contents |
Deploy, secrets, agent connect, tool → route table | |
Human SPA vs MCP matrix, gaps, voice workaround |
License
MIT (this package). Prism itself remains AGPL-3.0-only.
Hosted door
First-party Worker: https://prism-mcp.skyphusion.org (tag-gated deploy on prism-mcp-v*).
Self-host still supported via wrangler template.
This server cannot be deployed
Maintenance
Related MCP Connectors
Nifty's MCP server — exposes tasks, projects, messages, and files as tools for AI agents.
- UnifAPIOAuthcom.unifapi
Hosted MCP server for live public-data APIs and Skills for AI agents.
MCP server for AI agents to plan, verify, and deploy Cloudflare-native apps.
MCP server that lets AI assistants use all OneSchema features exposed via the public API.
Related MCP Servers
- AlicenseNot gradedqualityAmaintenanceProduction-ready MCP server providing RAG, hierarchical memory, and 8+ tools for AI agents via the Model Context Protocol.44Apache 2.0
- AlicenseNot gradedqualityDmaintenanceMCP server that provides tools to interact with the LiteLLM proxy API, enabling LLM completions, embeddings, image generation, and admin operations.8 npmMIT
- FlicenseNot gradedqualityDmaintenanceLocal MCP server that exposes fixed tools for GPT, Claude, and Gemini while routing to any OpenAI-compatible chat completions backend with independent configuration per target.1-
- FlicenseNot gradedqualityDmaintenanceHTTP MCP server for TaskFrame that enables AI agents to manage projects, tasks, comments, and documents through the TaskFrame API.-