Overview

The DevRev MCP Server implements a Model Context Protocol (MCP) gateway that lets you search and surface objects from a DevRev Knowledge Graph. It connects to DevRev APIs, ingests objects from multiple external sources, and exposes an HTTP-compatible MCP interface so downstream systems (search UIs, chat assistants, embedding services) can request contextual objects for a given model prompt or search query.

This server is useful when you need a single, consistent way to provide model context from heterogeneous systems. Instead of writing point-to-point integrations for each tool or dataset, the MCP server centralizes ingestion, normalization, and query logic for knowledge-graph objects and serves them via a small set of MCP-style endpoints.

Features

Integrates with DevRev APIs to read and sync knowledge graph objects.
Imports objects from multiple sources (e.g., third-party APIs, data exports, or internal systems).
Normalizes diverse objects into a unified schema consumable by model context requests.
Exposes HTTP endpoints following MCP conventions for fetch and search operations.
Pluggable configuration for source connectors, authentication, and filtering.
Lightweight server suitable for local testing, staging, or deployment behind an API gateway.

Installation / Configuration

Basic steps to get the repository and run locally:

Clone the repository:

git clone https://github.com/devrev/mcp-server.git
cd mcp-server

Install dependencies and build (example; follow repository-specific instructions if present):

# If the project uses Node.js
npm install
npm run build
npm start

# Or if the project is a compiled binary
# make build
# ./bin/mcp-server

Example environment variables (set before starting):

export DEVREV_API_TOKEN="your_devrev_api_token"
export MCP_PORT=8080
export MCP_CONFIG_PATH="./config/mcp-config.yaml"

Example minimal YAML configuration (config/mcp-config.yaml):

server:
  port: 8080
devrev:
  api_token: "${DEVREV_API_TOKEN}"
imports:
  - name: "tickets"
    type: "devrev_knowledge_graph"
    params:
      project: "my-project"
  - name: "external_docs"
    type: "http_export"
    params:
      url: "https://example.com/docs/export.json"
storage:
  path: "./data"
logging:
  level: "info"

Run via Docker (if an image is available or you build one):

docker build -t devrev/mcp-server .
docker run -e DEVREV_API_TOKEN="${DEVREV_API_TOKEN}" -p 8080:8080 devrev/mcp-server

Configuration tips:

Keep the DevRev API token secure (use secrets manager or environment injection in production).
Configure import schedules and filters to control which objects are indexed.
Use storage path or backing store settings to persist imported objects across restarts.

Available Resources

Source code and issues: https://github.com/devrev/mcp-server
Configuration examples: config/ (look for sample YAML/JSON in the repo)
Importer adapters: check the repo for supported connector types and example parameter sets
MCP endpoints: the server exposes HTTP endpoints to fetch context and perform searches (see API reference in the repo README or /openapi if provided)

API Examples

Search endpoint (example):

curl -X POST "http://localhost:8080/mcp/search" \
  -H "Authorization: Bearer $DEVREV_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "how to reset password",
    "limit": 10,
    "filters": { "object_type": ["ticket", "doc"] }
  }'

Example response shape (simplified):

{
  "results": [
    {
      "id": "ticket:1234",
      "type": "ticket",
      "title": "Resetting a forgotten password",
      "snippet": "To reset your password, go to ...",
      "source": "devrev",
      "score": 0.93
    }
  ],
  "total": 1
}

Fetch context for a model (MCP-style fetch):

curl "http://localhost:8080/mcp/fetch?object_id=ticket:1234" \
  -H "Authorization: Bearer $DEVREV_API_TOKEN"

Use Cases

Contextual answers in a chat assistant: When a customer asks a question, call /mcp/search to retrieve relevant tickets, docs, and knowledge-graph nodes to include in the model prompt so the assistant can produce grounded replies.
Unified search UI: Power a single search bar that returns results across DevRev items and external documentation by letting the MCP server normalize and rank results from multiple importers.
Automated triage and routing: Periodically import tickets and metadata, then run queries to identify high-priority or duplicate issues and surface them to workflows or dashboards.
Embeddings and retrieval-augmented generation (RAG): Use the server to export normalized objects for embedding pipelines or to provide retrieved context snippets at model inference time.

Best Practices

Limit the scope of imports on initial setup to a small dataset to validate mappings and ACL behavior.
Use filtering rules and metadata enrichment during ingestion so search and fetch responses contain the fields your models expect.
Monitor logging and enable verbose import logs while onboarding new sources; then switch to info-level for production.
Consider caching frequently requested context and securing endpoints via mTLS or an API gateway in production.

If you need more detailed API reference or connector samples, check the repository README and examples directory at https://github.com/devrev/mcp-server.

DevRev MCP Server for Knowledge Graph Search