signal-tools

Signal tools for coding agents: chats, messages, contacts, received files and sends.

The tools need Node.js 24 or later. docs/conventions.md describes how the tools are named and how they behave.

Install

Install from a release, never from the default branch: a release is a tag whose commit carries the built dist/, and the default branch does not. In the commands below, <version> is a release's version (1.2.3) and <sha> is the commit its tag names:

gh api repos/craft-agent/signal-tools/commits/v<version> --jq .sha

OpenCode

OpenCode V2 takes the package and its options in opencode.json:

{
  "plugins": [{ "package": "@craftsamo/signal-tools@<version>", "options": { "read_only": false } }],
  "permissions": [{ "action": "signal_*", "resource": "*", "effect": "ask" }]
}

OpenCode V1 takes the package as a string, or as a pair with its options:

{
  "plugin": [["@craftsamo/signal-tools@<version>", { "read_only": false }]],
  "permission": { "signal_*": "ask" }
}

OpenCode allows any action no rule names, so the plugin's request is granted silently unless a rule says ask. Read tools never ask, so the signal_* glob is safe.

V1 registers no skills. Unpack the package and copy its skills/* folders into ~/.config/opencode/skills/:

npm pack @craftsamo/signal-tools@<version>
tar -xzf craftsamo-signal-tools-<version>.tgz
mkdir -p ~/.config/opencode/skills
cp -r package/skills/* ~/.config/opencode/skills/
rm -rf package craftsamo-signal-tools-<version>.tgz

Claude Code

claude plugin marketplace add craft-agent/signal-tools#v<version>
claude plugin install signal-tools@craft-agent

Codex

codex plugin marketplace add craft-agent/signal-tools --ref v<version>
codex plugin add signal-tools@craft-agent

Codex turns - into _ in server names, so it shows the tools as mcp__signal_tools__signal_*. It starts the MCP server in the plugin's install directory, which is never the caller's working tree.

Grok Build

grok plugin install craft-agent/signal-tools@v<version> --trust

Grok starts the MCP server in the session's working directory.

Antigravity

Install from the tree of the release tag, never from the default branch, which has no built dist/; Antigravity clones the tag into ~/.gemini/config/plugins/signal-tools.

agy plugin install https://github.com/craft-agent/signal-tools/tree/v<version>

Hermes Agent

Hermes installs a commit, not a tag, and the plugin installs disabled:

hermes plugins install craft-agent/signal-tools --ref <sha>
hermes plugins enable signal-tools

If a Hermes gateway is already running, restart it (hermes gateway restart) to load the plugin.

Any other MCP host

Add the server to the host's MCP configuration. Flags go in args:

{
  "mcpServers": {
    "signal-tools": {
      "command": "npx",
      "args": ["-y", "@craftsamo/signal-tools@<version>", "--read-only"]
    }
  }
}

Related MCP server: telegram-mcp

Tools and skills

Each toolset comes with a skill that tells an agent how to use its tools.

Account (account)

It reads local state only: no daemon, no network, and no number, account id or key in its result. The skill is signal-account, shown as signal-tools:signal-account in Claude Code and signal-tools:account in Hermes.

Chats (chats)

It reads the account's local mirror only: no daemon, no network, and no number of the account's own, attachment id or key in its result. History starts when the device was linked. Times are local with their offset, a chat is always its id, a long result is paged with next, and everything it returns is data written by other people. The skill is signal-chats, shown as signal-tools:signal-chats in Claude Code and signal-tools:chats in Hermes.

Contacts (contacts)

Listing needs no daemon; checking numbers needs the running sync service. Neither writes the mirror. The skill is signal-contacts, shown as signal-tools:signal-contacts in Claude Code and signal-tools:contacts in Hermes.

Files (files)

The toolset is off by default: it writes to disk, opens archives and sends files. Turn it on with toolsets or tools. Files are sent only from the working directory (not when that is the home directory or a file system root) and the directories the send_dirs setting names; keys, secrets, programs and installers are refused, archives are checked entry by entry, and a file's sha256 can be given to be sure the file sent is the one approved. Sending needs the sync service running. Received files go to the download_dir setting, in a folder per message, private to the user (mode 700 and 600) and never executable; nothing is overwritten. Programs and installers are never saved, and an archive (zip, tar, gzipped tar) is saved only after its contents pass inspection, then listed; unpack: true unpacks it into <name>.unpacked/ beside it. A view-once message's files are not saved. On macOS every saved file carries the quarantine flag. Saving needs no daemon. The skill is signal-files, shown as signal-tools:signal-files in Claude Code and signal-tools:files in Hermes.

Sends (sends)

A chat is always its id; the user's own account id is Note to Self, and a group the user left is refused. A result means Signal accepted the message, with reached and not_reached for a group; any other failure after the request went out is uncertain: check the phone before sending again. The message is recorded in the mirror, and a record that fails is listed under errors without undoing the send. It needs the running sync service. The skill is signal-sends, shown as signal-tools:signal-sends in Claude Code and signal-tools:sends in Hermes.

Actions (actions)

The toolset is off by default: the actions change what others have already read. Turn it on with toolsets or tools. Editing and deleting take only the user's own messages, within 24 hours of sending, and a message is edited at most 10 times; only a text message can be edited. A reaction goes to any message, one per person, so a new one replaces the user's earlier one. Outcomes follow the sends': a result means Signal accepted it, with reached and not_reached for a group, and any other failure after the request went out is uncertain: check the phone before repeating it. Each is recorded in the mirror as the sync service would record it from another device, and a record that fails is listed under errors. No tool sends read receipts or typing indicators. It needs the running sync service. The skill is signal-actions, shown as signal-tools:signal-actions in Claude Code and signal-tools:actions in Hermes.

Adding a tool

1. Write the spec in src/core/<domain>/<verb>.ts with defineTool: an ID <prefix>_<domain>_<verb>, a title and description, the Zod argument and output schemas, its toolset and write class, and execute(args, context). A write tool also gives targets(args, context): what the call writes to, as a permission rule should name it.

2. Test it beside the spec (<verb>.test.ts), calling it through invoke as the entry points do. A tool that runs a program is tested against a fake of that program on PATH; no test reaches a real service.

3. List it in src/tools.ts. A new toolset goes into toolsets in src/project.ts (and defaultToolsets when it is on by default).

4. Add or update the toolset's skill, skills/<prefix>-<toolset>/SKILL.md.

5. Run pnpm run build && pnpm run manifests to list the tool in plugin.yaml, then describe it under "Tools and skills".

docs/conventions.md has the rules each step follows.

Reads and writes

Every read tool is named <prefix>_<domain>_read. Reads can run unasked; writes are confirmed by the harness.

Where a write is asked about, and where it is not:

• OpenCode grants any action no rule names, so a write runs unasked unless a rule asks for it: add the signal_* rule shown under OpenCode.

• Codex under codex exec, and Grok Build under grok -p, refuse a write they cannot ask about, unless it is allowed beforehand.

• Grok Build calls the plugin's tools through its use_tool tool, and the agent often names a tool without its server first. Grok checks the rules against the name as given, so the server-qualified rule alone does not match that first call. The second rule matches the short name; a write still matches neither.

• Antigravity asks before any MCP tool no rule allows, reads included, and names the plugin's server signal-tools_signal-tools (plugin, then server). agy -p refuses a tool it cannot ask about.

• Hermes hermes chat -q refuses a write it cannot ask about; hermes -z and --yolo skip approval, so writes then run unasked.

Settings

Every entry point takes the same settings. A flag wins over its environment variable, and an option or setting wins over the variable too.

Lists are comma-separated in a flag or a variable. Where each harness reads them:

• OpenCode takes the plugin options shown above, with lists as arrays, read_only as a boolean, state_dir and download_dir as paths and send_dirs as an array of paths, and falls back to the variables. V2 runs plugins in its background service, which keeps the environment it started with: a variable set where opencode run starts has no effect until the service is restarted in that environment (opencode service restart). V1 runs plugins in its own process.

• Hermes takes plugins.entries.signal-tools.settings in config.yaml:

  plugins:
    entries:
      signal-tools:
        settings:
          read_only: false
          exclude_tools: []

  The plugin passes its own environment to the command line it runs, so the variables work too. A running gateway keeps the environment it started with; restart it (hermes gateway restart) after changing one.

• Claude Code, Grok Build and Antigravity start the plugin's server without flags and pass it the client's environment, so set the variables in the environment the harness runs in.

• Codex passes a plugin's MCP server only a few basic variables (such as PATH and HOME), so the variables have no effect there. To change settings, or to give the server a variable it needs, add the server under [mcp_servers] in Codex's config.toml instead of using the plugin: flags go in args, env_vars names the variables to forward from Codex's environment, and env sets fixed values.

  [mcp_servers.signal-tools]
  command = "npx"
  args = ["-y", "@craftsamo/signal-tools@<version>", "--read-only"]
  env_vars = ["SIGNAL_TOOLS_STATE_DIR"]

Accounts and the sync service

The tools use a Signal account as a linked device of the user's phone, the way Signal Desktop does, through signal-cli, which you install yourself. They never read another client's data.

• Each linked device has a label, 1 to 24 lowercase letters, digits and _ (default when linking names none). A tool's account argument picks the label; leave it out when only one device is linked.

• Each label keeps its state in one directory, <state_dir>/<label>/ (mode 700): signal-cli's data and keys, the daemon's socket, the mirror of received messages, and the copies of files being sent. state_dir defaults to $XDG_STATE_HOME/signal-tools, or ~/.local/state/signal-tools. Keep it out of backups you share: the keys read and send as you.

• A sync service receives for each label and keeps its mirror, and the read tools read only the mirror. History starts when the device is linked; messages that arrive while the service is down wait on Signal's server, for as long as Signal keeps undelivered messages.

• Signal unlinks a device unused for 45 days, and every linked device when the phone has been offline for 30; a running service keeps the device in use. An unlinked label is linked again from the phone.

• signal-cli releases age out. A release more than about three months old may stop working with Signal's servers, so keep it updated.

Linking a device and running the sync service

The tools are tested against signal-cli 0.14. Install it first: on macOS brew install signal-cli; elsewhere a release (the JVM build needs Java 25 or later). macOS and Linux are supported; Windows is #11.

# Link: on the phone, Signal > Settings > Linked devices > Link new device,
# then scan the code this prints.
signal-tools-cli link --account work

# Start receiving at logon, and restart after a failure.
signal-tools-cli service install --account work
signal-tools-cli service status --account work

--account may be left out when only one label exists (link names a new one default), and every command takes --state-dir like the tools.

• The service records this package's paths, so run service install again after updating the package. On macOS its log is ~/Library/Logs/io.github.craft-agent.signal-tools/<label>.log; on Linux, journalctl --user -u <unit> as service status shows, and loginctl enable-linger keeps it running while you are logged out.

• A phone takes at most five linked devices, and each label is one of them.

• Removing a label does not unlink it on the phone: a linked device cannot take itself off, so remove it under Settings > Linked devices as well.

• To link a label again after Signal unlinked it, run link again: the old keys are moved aside and the history is kept, as long as the phone holds the same account. A different account gets a label of its own.

• signal_account_read tells an agent which of these steps a label needs.

Layout

• src/project.ts — the package's name, tool ID prefix, setting prefix and toolsets.

• src/core/ — the tool contract and the tools, free of any harness and of the project's names.

• src/tools.ts — the list of every tool, which each entry point registers.

• src/mcp/ — the MCP server, built to dist/mcp.js and run by signal-tools.

• src/opencode/ — the OpenCode plugin, built to dist/opencode.js.

• src/cli/ — the command line the Hermes plugin calls, built to dist/cli.js and run by signal-tools-cli.

• src/hermes/, plugin.yaml and __init__.py — the Hermes plugin.

• skills/ — one skill per toolset, skills/<name>/SKILL.md, which every harness that takes skills loads from the package or plugin root.

• plugin.json, mcp.json, .mcp.json, .claude-plugin/, .agents/plugins/, .antigravity-plugin/ and mcp_config.json — the plugin manifests of the other harnesses. The root .mcp.json is the configuration Grok Build reads (with ${CLAUDE_PLUGIN_ROOT}, the variable both Grok Build and Claude Code expand); when working on this repository in Claude Code, decline the project MCP server it offers from that file.

How a release is made is in RELEASING.md.

License

MIT