agy-discord-mcp
Bridges a Discord bot with the agy CLI, enabling inbound messages to be relayed to agy and allowing agy to interact with Discord through MCP tools, including sending messages and files.
README
agy-discord-mcp
Discord bridge and MCP tools for the agy (Antigravity) CLI. It connects a Discord bot to agy in two ways:
botrelay — inbound Discord messages are sent toagy --printand agy's reply is posted back (text only).mcpmode — exposes Discord tools (reply, send_message with file attachments, react, fetch history, download attachments, …) to an interactive agy session over stdio, so agy can talk to Discord itself — including posting generated images.
It is a sibling of codex-discord-mcp, ported to agy. Because agy --print writes a clean response straight to stdout (no JSON to parse) and agy saves generated images as real files on disk, the bridge is a little simpler than the Codex original.
Requirements
- Node.js >= 20
- The
agy(Antigravity) CLI installed and authenticated (check withagy --version) - A Discord bot token (with the Message Content intent enabled)
Install
npm install -g agy-discord-mcp
agy-discord-mcp doctor
Or run it without a global install:
npx agy-discord-mcp@latest doctor
For local development from this repository:
npm ci
npm run build
npm link
Configure
Store your Discord bot token (written to ~/.agy/discord/.env, mode 600):
agy-discord-mcp configure <bot-token>
# or interactively:
agy-discord-mcp init
Invite the bot and check local status:
agy-discord-mcp invite-url <discord-client-id>
agy-discord-mcp doctor
Access control
Inbound access is allowlist/pairing based and managed only from the CLI — never from Discord messages:
agy-discord-mcp access show
agy-discord-mcp access policy <pairing|allowlist|disabled>
agy-discord-mcp access allow-user <discord-user-id>
agy-discord-mcp access allow-channel <channel-id> [--no-mention] [--allow-user <id>...]
agy-discord-mcp access pair <code> # approve a DM pairing code
Under the default pairing policy, the first DM from an unknown user returns a one-time code; run access pair <code> on the host to approve them.
bot relay mode
agy-discord-mcp bot
Each allowed message becomes agy --print "<prompt>" run in AGY_WORKDIR, and the trimmed stdout is posted back. Set AGY_RESUME_BY_CHANNEL=true to keep a per-channel agy conversation (resumed via --conversation <id>, detected from agy's conversations directory).
Files come back in relay mode too — not just images. The relay adds the output dir (default ~/agy_images, set by AGY_DISCORD_GENERATED_IMAGES_DIR) to agy's workspace and tells agy to save any user-facing file there; any deliverable agy writes during the turn — images plus common document/data/archive types (md, html, pdf, csv, json, zip, docx, …; code/temp files are ignored) — is detected and attached to the Discord reply automatically (≤25 MB each, up to 10). This needs no MCP server. If other agy workloads also write to ~/agy_images (e.g. image-gen crons), point the bot at a dedicated dir via AGY_DISCORD_GENERATED_IMAGES_DIR so their output isn't picked up.
agy is launched through its wrapper, which auto-injects
--dangerously-skip-permissions— every tool call is auto-approved. Treat Discord input as untrusted: run in an isolated workspace, setAGY_SANDBOX=1, or acknowledge the risk withAGY_DISCORD_ASSUME_YES=true.
mcp mode (agy drives Discord)
Caveat (current agy): registering an MCP server in agy's settings makes a headless
agy --printhang on a first-use, interactive "trust this MCP server?" prompt — there is no config key (themcpServersschema has notrustfield) oragy mcpsubcommand to pre-approve it. Since that global setting is read by every agy invocation, it also breaks the relay's ownagy --printcalls and any other headless agy users. Enable this only for a dedicated interactive agy session that can answer the prompt; for normal use prefer relay mode above, which already attaches images.
Register the MCP server by merging a mcpServers block into ~/.gemini/settings.json (or ~/.gemini/antigravity-cli/settings.json). Generate it with:
agy-discord-mcp print-config # node + absolute path form
agy-discord-mcp print-config --npx # npx form
Example (agy-mcp-config.example.json):
{
"mcpServers": {
"discord": {
"command": "node",
"args": ["/path/to/agy-discord-mcp/dist/cli.js", "mcp"],
"timeout": 60000,
"trust": true
}
}
}
Then run agy normally; it can call the Discord tools below. Inbound Discord messages are queued — poll with list_pending_messages, reply, then mark_message_handled.
Sending images / files
agy writes generated images as real files (its native generate_image tool, or the agy-image skill, save to a path you choose — default ~/agy_images). Pass that absolute path in the files array of reply/send_message. Files must live under an allowed attachment root (see AGY_DISCORD_ATTACHMENT_ROOTS); the agy image dir and the bridge inbox are always allowed.
Tools
| tool | purpose |
|---|---|
reply / send_message |
post text + optional file attachments |
react |
add an emoji reaction |
edit_message |
edit a message the bot sent |
fetch_messages |
recent channel history (Discord bot search is unavailable) |
download_attachment |
save a message's attachments to the inbox |
latest_generated_images |
newest images under the agy image dir |
list_pending_messages / mark_message_handled |
inbound message queue |
bridge_status |
state dir, queue counts, Discord login status |
Environment
| var | default | purpose |
|---|---|---|
DISCORD_BOT_TOKEN |
— | bot token (usually stored in the state .env) |
AGY_DISCORD_STATE_DIR |
~/.agy/discord |
state directory |
AGY_COMMAND |
agy |
agy executable |
AGY_WORKDIR |
cwd | working directory for relay runs |
AGY_MODEL |
— | --model |
AGY_SANDBOX |
false |
pass --sandbox (terminal restrictions) |
AGY_RESUME_BY_CHANNEL |
false |
resume a per-channel conversation |
AGY_TIMEOUT_MS |
900000 |
relay run budget (also sets agy's --print-timeout) |
AGY_EXTRA_ARGS |
— | extra agy args (shell-style string or JSON array) |
AGY_CONVERSATIONS_DIR |
~/.gemini/antigravity-cli/conversations |
where agy stores conversation .db files |
AGY_DISCORD_ATTACHMENT_ROOTS |
cwd + AGY_WORKDIR + inbox |
allowed upload roots (os-delimiter separated) |
AGY_DISCORD_GENERATED_IMAGES_DIR |
~/agy_images |
output dir the relay tells agy to save deliverables in and scans to auto-attach (always attachable) |
AGY_DISCORD_ASSUME_YES |
false |
suppress the unsafe-mode warning |
Development
npm run typecheck
npm test
npm run build
npm run dev:bot # tsx src/cli.ts bot
npm run dev:mcp # tsx src/cli.ts mcp
Publishing
GitHub Releases publish to npm through .github/workflows/release.yml.
- Make sure npm trusted publishing is configured for
Openclaw-Metis/agy-discord-mcpand therelease.ymlworkflow. - Bump
package.jsonandpackage-lock.jsonto the release version and commit the change. - Create a GitHub Release whose tag is
vplus the package version, for examplev0.1.0.
The release workflow checks that the tag matches package.json, runs typecheck, tests, and build, then publishes to npm with provenance. Normal releases use the npm latest dist-tag; GitHub pre-releases use next.
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.