GPS
Provides sub-millisecond access to organizational and engineering data by pre-materializing data into a read-only SQLite database. Enables agents to query people, teams, issues, features, and governance documents without upstream latency or auth.
README
GPS - Global Positioning System MCP Server
A read-only caching tier that gives agents and humans sub-millisecond access to org and engineering data. GPS materializes people, teams, issues, features, release schedules, component mappings, and governance documents into a SQLite database and serves them over MCP — optimized for high-frequency agent queries with zero upstream latency.
Why GPS exists: Agents need fast, reliable access to org and engineering data. Querying live APIs (Jira, Confluence, HR systems) on every turn is slow, rate-limited, and fragile. GPS pre-materializes everything into a single SQLite file, opens it read-only, and serves structured results in microseconds. No auth, no rate limits, no network dependencies at query time.
5-Minute Quickstart
# 1. Clone and enter
git clone https://github.com/YOUR-ORG/gps.git && cd gps
# 2. Install uv (if needed)
curl -LsSf https://astral.sh/uv/install.sh | sh
# 3. Copy env and customize org mapping
cp .env.example .env
# 4. Place your data files in data/
# - An XLSX org spreadsheet (any .xlsx file)
# - CSV exports: issues, features, releases, etc.
# (Example data ships in data/acme-* for testing)
# 5. Build the database
uv run scripts/build_db.py --force
# 6. Run tests
scripts/test.sh
# 7. Start the MCP server
uv run mcp_server.py # stdio (for Claude Code / ACP)
uv run mcp_server.py --http # HTTP on :8000 (for shared deployments)
Architecture
graph TD
subgraph Sources["Data Sources (refreshed periodically)"]
XLSX["XLSX\nOrg Spreadsheet"]
CSV["CSV\nIssues, Features,\nReleases"]
GOV["Markdown / PDF\nGovernance Docs"]
end
ETL["build_db.py\n(ETL Pipeline)"]
DB[("gps.db\nSQLite — read-only cache\nsub-ms queries")]
subgraph Server["MCP Server (no auth required)"]
MCP["mcp_server.py"]
TOOLS["9 Tools\nlookup_person, search_issues,\nrelease_risk_summary, ..."]
RES["2 Resources\ngps://schema\ngps://catalog"]
end
subgraph Clients["Clients"]
ACP["ACP Sessions\n(stdio, every pod)"]
CC["Claude Code\n(stdio, local)"]
HTTP["Shared Deployment\n(HTTP :8000)"]
end
XLSX --> ETL
CSV --> ETL
GOV --> ETL
ETL --> DB
DB --> MCP
MCP --- TOOLS
MCP --- RES
MCP -->|stdio| ACP
MCP -->|stdio| CC
MCP -->|streamable-http| HTTP
style DB fill:#e8f4f8,stroke:#2196F3,stroke-width:2px
style MCP fill:#fff3e0,stroke:#FF9800,stroke-width:2px
style ETL fill:#f3e5f5,stroke:#9C27B0,stroke-width:2px
How it works
- ETL pipeline (
scripts/build_db.py) loads all sources into a single SQLite database — runs periodically, not per-query - MCP server (
mcp_server.py) opens the database read-only with mmap, 64MB cache, and memory-backed temp store — tuned for agent query patterns - No auth required — the database contains read-only organizational data; agents connect directly via stdio or HTTP
- The LLM never touches upstream data sources directly — clean security boundary
Client-Server Interaction
sequenceDiagram
participant Agent as Agent / Human
participant MCP as GPS MCP Server
participant DB as gps.db (SQLite)
Note over DB: Pre-built by ETL pipeline<br/>Opened read-only at startup
Agent->>MCP: tool call: lookup_person(name="Dana")
MCP->>DB: SELECT ... FROM person WHERE name LIKE '%Dana%'
DB-->>MCP: rows (sub-ms)
MCP-->>Agent: JSON {results: [...], count: 1}
Agent->>MCP: tool call: search_issues(status="In Progress", component="api")
MCP->>DB: SELECT ... FROM jira_issue WHERE status LIKE ... AND EXISTS(...)
DB-->>MCP: rows (sub-ms)
MCP-->>Agent: JSON {issues: [...], count: 12}
Agent->>MCP: tool call: release_risk_summary()
MCP->>DB: SELECT ... FROM release_milestone
MCP->>DB: SELECT ... FROM feature_release
MCP->>DB: SELECT ... FROM feature WHERE feature_id IN (...)
DB-->>MCP: rows
MCP-->>Agent: JSON {releases: [...], assessed_on: "2026-03-20"}
Agent->>MCP: resource: gps://schema
MCP->>DB: SELECT sql FROM sqlite_master
DB-->>MCP: DDL statements
MCP-->>Agent: Full schema with row counts
Note over Agent,DB: All queries are read-only<br/>No writes, no auth, no rate limits
MCP Tools
| Tool | Description |
|---|---|
lookup_person |
Find people by name, email, or user ID (partial match) |
list_team_members |
List all members of a scrum team with roles and components |
search_issues |
Search issues by status, priority, assignee, component, label, or keyword |
get_feature_status |
Get feature details: progress, RICE score, releases, components, teams |
release_risk_summary |
Assess release risk — flags features under 80% complete near milestones |
list_documents |
List governance documents with table of contents |
get_document |
Retrieve full governance document content by ID |
get_document_section |
Retrieve a specific section by fuzzy heading match |
get_gps_version |
Return GPS version and build metadata |
MCP Resources
| URI | Description |
|---|---|
gps://schema |
Full database DDL with row counts — agents should read this first |
gps://catalog |
Data source inventory (DATA_CATALOG.yaml) |
Wiring GPS into ACP Sessions
GPS runs as a sidecar MCP in every ACP pod — no auth needed. The recommended approach is adding it to the runner's managed settings:
{
"mcpServers": {
"gps": {
"command": "uv",
"args": ["run", "--script", "/app/gps/mcp_server.py"]
}
}
}
Bake mcp_server.py + data/gps.db + VERSION into the runner image or mount via shared volume. For init container and HTTP sidecar patterns, see docs/DEPLOYMENT.md.
Project Structure
mcp_server.py MCP server (stdio default, --http for HTTP)
scripts/
build_db.py ETL pipeline — materializes gps.db from source files
test.sh Test suite (lint, build, integrity, schema diff)
data/
acme-* Example data files (tracked)
*.csv, *.xlsx, *.db User data files (gitignored)
deploy/
deploy.sh Build, apply, status, logs automation
k8s/ Kubernetes/OpenShift manifests (kustomize)
docs/
adr/ Architecture Decision Records
DEPLOYMENT.md Deployment guide (local, container, k8s, ACP)
CUSTOMIZATION.md Customization guide (env vars, adding sources)
SCHEMA.md Database schema reference (ER diagram, tables, views)
governance/ Policy documents (auto-loaded into DB)
Containerfile Container image build
.env.example Configuration template
.mcp.json Claude Code MCP server config
Configuration
GPS is configured via environment variables (see .env.example):
GPS_TAB_ORG_MAP— JSON mapping of XLSX tab names to[org_key, org_name]pairsGPS_JIRA_SCRUM_REF_TAB— XLSX tab name for Jira-to-Scrum-team mappings
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.