Delphi is the context engine that backs Synthetic Sciences. It is a self-hosted MCP server that gives AI agents deep context through semantic search across three domains.
DomainWhat it does
Code repositoriesIndex GitHub repos, search code semantically, find functions and classes, analyze architecture
Research papersIndex arXiv papers or PDFs, extract citations and equations, generate reports
HuggingFace datasetsIndex dataset cards, search metadata
Everything runs on your machine. PostgreSQL for storage. Pick local sentence-transformers (no API keys), or wire up a hosted embeddings provider. Open source under Apache 2.0. The repo lives at github.com/synthetic-sciences/delhpi.

​
Quick start

npx @synsci/delphi
The installer asks two questions: which agent host to wire up (Claude Code, Cursor, Windsurf, Claude Desktop), and which embeddings provider. Then it pulls the source, spins up the Docker stack, mints an API key, and writes the MCP config for the tools you have installed. When it finishes, restart your AI tool. Delphi shows up as an MCP server.

​
Daily commands

delphi              # open the dashboard (boots the stack if it's down)
delphi status       # check health and container state
delphi logs -f      # tail logs
delphi stop         # tear it down
delphi uninstall    # remove containers and data volume
The dashboard runs at localhost:3000. The API runs at localhost:8742.

​
How Delphi fits with Thesis

Thesis stores your research as a graph of hypotheses, experiments, sources, and decisions. Delphi is the layer underneath: it indexes the code, papers, and datasets you point it at, and exposes that index over MCP so any compatible agent can search it. Use them together when your research draws on a body of code or literature larger than what a single chat context can hold. Thesis tracks your reasoning. Delphi gives the agent the context to reason with.

​
Manual install

For contributors or anyone running a fork: 
git clone https://github.com/synthetic-sciences/delhpi.git
cd delhpi
cp env.example .env       # set SERVER_SECRET and SYSTEM_PASSWORD
./scripts/launch_app.sh   # or: docker compose up --build
Requirements: Docker Desktop or docker compose v2, plus git.

​
Wiring it into your agent host manually

Once you have an API key (from the dashboard at /api-keys or via npx @synsci/delphi init), add the MCP server to your client’s config: 
{
  "mcpServers": {
    "synsci-delphi": {
      "command": "uvx",
      "args": ["synsc-context-proxy"],
      "env": {
        "SYNSC_API_KEY": "your-api-key",
        "SYNSC_API_URL": "http://localhost:8742"
      }
    }
  }
}
HostConfig path
Cursor~/.cursor/mcp.json
Windsurf~/.codeium/windsurf/mcp_config.json
Claude Desktop~/Library/Application Support/Claude/claude_desktop_config.json
Claude Codeclaude mcp add --scope user synsci-delphi -- uvx synsc-context-proxy