Model Context Protocol
Context teleportation.
Work in any AI. Your knowledge follows. Penumbra is a unified semantic stack—a canonical source of truth that any AI can read from and write to.
The Problem
Your context is trapped.
Every AI conversation dies when the session ends. Switch from Claude to ChatGPT? Start over. Build a custom agent? Manually re-inject everything. Your AI work is fragmented across platforms, forever siloed.
Session amnesia
Chat logs die. The brilliant synthesis you did last week? Gone. The context you built up? Evaporated.
Platform lock-in
Your Claude context doesn't exist in ChatGPT. Your GPT threads don't exist in Claude. Siloed by default.
Manual re-injection
Building agents? You're copy-pasting context, writing custom loaders, rebuilding state every time.
The Shift
What if context was portable?
Penumbra is the canonical source of truth for your knowledge. Any AI can pull your workspace, work with full context, and push updates back. Your context travels with you.
Unified Semantic Stack
Claude
ChatGPT
Your Agent
Penumbra Workspace
Your knowledge graph • Structured entities • Full contextKnowledge graph • Entities • Context
Cursor
Penumbra App
MCP Client
The Flow
Capture. Pull. Work. Push.
01
Capture
Work in any AI. Valuable context gets captured to your Penumbra workspace as structured knowledge—not chat logs, but entities and relationships.
02
Pull
Switch to a different AI. It pulls your workspace context via MCP. Full knowledge graph, ready to go. No re-explaining.
03
Work
Reason, synthesize, create. The AI has everything it needs—your terms, your entities, your prior decisions.
04
Push
New insights flow back to Penumbra. Your workspace evolves. The next AI you use starts with everything.
Use Cases
What you can build.
Cross-platform workflows
Start research in ChatGPT, synthesize in Claude, review in Penumbra. Your context stays coherent across every transition.
Multi-agent orchestration
Your sales agent, research agent, and writing agent all share one knowledge base. No manual context passing.
Session capture
Every valuable AI conversation becomes persistent, queryable knowledge. Nothing dies in a chat log.
Team knowledge sync
Team members work in different AIs but share one canonical workspace. Context compounds across the org.
Technical
MCP-native. API-first.
Penumbra speaks Model Context Protocol natively. Connect Claude Desktop, Claude Code, or any MCP-compatible client. For everything else, there's the REST API.
- →MCP server for Claude Desktop & Claude Code
- →REST API for custom integrations
- →Webhooks for real-time sync
- →Structured workspaces, not text blobs
- →Entity-aware retrieval with relationships
- →Read and write—context flows both ways
Your context is waiting.
3 projects · 847 entities · 1,204 relationships
Why Penumbra
State-of-the-art Context Engine.
Traditional RAG
- ×Chunks of text by proximity
- ×No entity awareness
- ×Read-only retrieval
- ×Context window limited
- ×Rebuild for each platform
Penumbra MCP
- →Structured entities with relationships
- →Your terms, your frameworks
- →Read and write—bidirectional
- →Intelligent context selection
- →One stack, every platform
FAQ
Technical questions.
Model Context Protocol is an open standard created by Anthropic for connecting AI models to external data and tools. It's how Claude Desktop, Claude Code, and other clients can securely access your Penumbra workspace.
Ready to Build?
Your context. Every AI.
Get API access and start building with context teleportation.