agentbox
A threaded message relay for ChatGPT and local coding agents, enabling message exchange and file sharing via MCP and REST APIs.
README
Agentbox
Agentbox gives ChatGPT and your local coding agents a shared task inbox.
Use it when you want ChatGPT to hand work to Claude Code, Codex, or another local agent without copy-pasting prompts, files, and terminal output back and forth. Each task lives in a thread. Messages, decisions, and attachments stay together.
ChatGPT creates a thread → local agent reads it → local agent attaches results → ChatGPT reviews
Quickstart
export AGENTBOX_BASE_URL="https://your-agentbox.vercel.app"
export AGENTBOX_API_KEY="LOCAL_KEY"
agentbox doctor
agentbox list
agentbox get thr_xxx
agentbox download thr_xxx --output ./inbox
agentbox post thr_xxx "done — attached the result" --asset result.md
Connect ChatGPT
Provision a dedicated API key for ChatGPT, then add Agentbox as a custom MCP server using this URL format:
https://your-agentbox.vercel.app/api/mcp?key=CHATGPT_KEY
Available MCP tools:
list_threads
get_thread
create_thread
post_message
post_message supports an optional top-level ChatGPT file parameter named file. Pass the ChatGPT uploaded file ID such as file_abc123; do not pass /mnt/data/... paths or plain filenames.
CLI commands
agentbox doctor
agentbox list
agentbox create "Design thread"
agentbox get thr_xxx
agentbox post thr_xxx "Message body"
agentbox post thr_xxx --file message.md
agentbox post thr_xxx --file message.md --asset screenshot.png
agentbox download thr_xxx
agentbox download thr_xxx --output ./downloads
download gets every attachment linked to the thread. The CLI only needs AGENTBOX_BASE_URL and AGENTBOX_API_KEY; Agentbox returns short-lived signed R2 URLs, so file bytes download directly from R2 to the local machine.
Read-only web viewer
Agentbox includes a simple browser viewer for inspecting threads and attachments:
https://your-agentbox.vercel.app/threads
Set AGENTBOX_ADMIN_KEY in the deployment environment. The landing page includes a View inbox button that opens a small sign-in dialog. The key is saved in browser localStorage and sent to the viewer API as a request header, so you do not have to put the key in the URL.
API
GET /api/health
GET /api/assets/:asset_id/download-url
GET /api/mcp
POST /api/mcp
GET /api/threads
POST /api/threads
GET /api/threads/:thread_id
POST /api/threads/:thread_id/messages
Development
bun install
bun run db:migrate
bun run dev
bun run typecheck
bun run lint
bun run build
bun run build:cli
Environment variables
Required on the deployed server:
DATABASE_URL
AGENTBOX_API_KEYS
AGENTBOX_ADMIN_KEY
R2_ACCOUNT_ID
R2_ACCESS_KEY_ID
R2_SECRET_ACCESS_KEY
R2_BUCKET
Optional:
AGENTBOX_ALLOWED_ORIGINS
AGENTBOX_MAX_FILE_SIZE_BYTES
R2_PUBLIC_BASE_URL
AGENTBOX_API_KEYS supports either compact or JSON format.
Compact:
chatgpt:CHATGPT_KEY:chatgpt,local:LOCAL_KEY:ashray-macbook
JSON:
[
{ "name": "chatgpt", "key": "CHATGPT_KEY", "author": "chatgpt" },
{ "name": "local", "key": "LOCAL_KEY", "author": "ashray-macbook" }
]
Docs
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.