devspace-memory-bridge
A non-invasive MCP bridge that makes DevSpace search tools memory-aware by augmenting search/list tool responses with codebase memory context.
README
devspace-memory-bridge
A non-invasive MCP bridge that makes DevSpace search tools memory-aware.
devspace-memory-bridge sits in front of a normal DevSpace server and augments selected search/list tool responses with structural context from codebase-memory-mcp. The model still calls ordinary DevSpace-style tools such as grep, glob, and ls; the bridge enriches the response locally before it reaches the client.
ChatGPT / MCP client
-> devspace-memory-bridge
-> upstream DevSpace
-> codebase-memory-mcp CLI
Why this exists
Some coding agents support hooks that can automatically inject codebase context when the agent searches files. Other MCP clients do not expose that kind of hook surface. This project explores a practical middle ground: intercept only the DevSpace search/list tools at the local boundary, call the real DevSpace tool, call codebase-memory-mcp, and return one combined tool result.
The goal is not to replace DevSpace or codebase-memory-mcp. The goal is to connect them without patching either upstream project.
What it does
Implemented:
- Records
workspaceId -> rootfromopen_workspaceresults. - Passes unknown and non-memory tools through unchanged.
- Augments
grep/grep_fileswithsearch_graphcontext. - Augments
glob/find_fileswithsearch_graphcontext. - Augments
ls/list_directorywith lightweight architecture context. - Transforms JSON-RPC JSON responses.
- Transforms request-scoped MCP SSE responses from
POST /mcp. - Streams standalone
GET /mcpSSE responses through unchanged. - Fails open by default: if memory lookup fails, the original DevSpace result is returned.
Intentionally not implemented yet:
bashcommand interception.read,edit, orwriteaugmentation.- A full DevSpace replacement.
- A general-purpose MCP middleware framework.
Not yet validated end-to-end:
- ChatGPT Apps OAuth/resource-server registration through the bridge.
- Long-lived production deployment behind a public HTTPS tunnel.
Tool behavior
The bridge treats upstream DevSpace as the source of truth. A normal request flow is:
1. Client calls grep(pattern="TurnController")
2. Bridge forwards the call to upstream DevSpace
3. DevSpace returns normal grep results
4. Bridge extracts symbol hints from the search pattern
5. Bridge calls codebase-memory-mcp search_graph
6. Bridge appends "Codebase Memory Context" to the tool result
7. Client receives what still looks like a grep result
If step 5 fails, step 6 is skipped and the original DevSpace response is returned.
Requirements
- Node.js 22 or newer.
- A running DevSpace server.
codebase-memory-mcpavailable as a CLI command, or a compatible wrapper command.- Upstream DevSpace should run with dedicated search/list tools enabled:
DEVSPACE_TOOL_MODE=full node /path/to/devspace/dist/cli.js serve
Quick start
Start upstream DevSpace:
DEVSPACE_TOOL_MODE=full node /home/heppo/DEV/devspace/dist/cli.js serve
Start the bridge:
cd /home/heppo/DEV/devspace-memory-bridge
DEVSPACE_UPSTREAM_URL=http://127.0.0.1:7676 \
DEVSPACE_MEMORY_BRIDGE_PORT=7777 \
DEVSPACE_MEMORY_COMMAND=codebase-memory-mcp \
node src/index.js
Point the MCP client at the bridge instead of upstream DevSpace.
Configuration
| Variable | Default | Purpose |
|---|---|---|
DEVSPACE_UPSTREAM_URL |
http://127.0.0.1:7676 |
Real DevSpace server URL. |
DEVSPACE_MEMORY_BRIDGE_HOST |
127.0.0.1 |
Bridge bind host. |
DEVSPACE_MEMORY_BRIDGE_PORT |
7777 |
Bridge bind port. |
DEVSPACE_MEMORY_COMMAND |
codebase-memory-mcp |
CLI command used for memory lookups. |
DEVSPACE_MEMORY_AUGMENT_TOOLS |
grep,glob,ls |
Tools to augment. |
DEVSPACE_MEMORY_FAIL_OPEN |
true |
Return normal DevSpace results if memory lookup fails. |
DEVSPACE_MEMORY_MAX_CONTEXT_CHARS |
6000 |
Max appended memory context length. |
Tests
npm test
npm run check
Current test coverage focuses on the bridge logic rather than a live ChatGPT/DevSpace OAuth round trip:
- memory context formatting
- symbol hint extraction
- JSON-RPC tool response transformation
- fail-open behavior
- non-memory tool pass-through
- request-scoped SSE transformation
Security model
This bridge sits in front of a local coding workspace. Treat it as sensitive infrastructure.
- Bind to localhost by default.
- Put authentication and HTTPS at the outer boundary if exposed remotely.
- Do not expose it directly to the public internet.
- Keep fail-open enabled unless you explicitly want memory failures to break tool calls.
- Review upstream DevSpace permissions and allowed roots before use.
Design rules
- Do not patch DevSpace upstream.
- Do not patch
codebase-memory-mcpupstream. - Treat DevSpace as the source of truth for filesystem operations.
- Only append memory context after the upstream DevSpace call succeeds.
- Unknown tools must pass through unchanged.
- Memory failures must not break local coding tools.
Status
Prototype. The core transformation behavior is implemented and tested, but production ChatGPT Apps registration through the bridge still needs end-to-end validation.
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.