> ## Documentation Index
> Fetch the complete documentation index at: https://docs.heysage.me/llms.txt
> Use this file to discover all available pages before exploring further.

# Telegram Agent

> Approve, reject, analyze, and configure your vault from Telegram.

Sage's conversational layer lives in Telegram. When a proposal needs your attention, the bot messages you; you respond with a tap or a command. Behind the scenes, a [Nanobot](https://github.com/nanobot-ai/nanobot) skill pack routes your messages to the Sage API.

The agent is **conversational only** — every deterministic decision (scoring, thresholds, execution) lives on the server. The agent calls endpoints and reports the actual result.

## Connecting the Bot

Open the Sage bot in Telegram and send `/start`. The bot pings the server (`POST /bot-ping`), which links your chat to your vault. Once connected, screening notifications and your replies flow through.

## Notification Buttons

A REVIEW or BLOCK message includes three buttons:

| Button          | Action                                                                          |
| --------------- | ------------------------------------------------------------------------------- |
| ✅ approve       | Co-sign and execute the proposal, then update the message with the tx signature |
| ❌ reject        | Mark the proposal rejected and update the message                               |
| 🔎 deep-analyze | Run live GoPlus + Rugcheck + threat-intel analysis and report findings          |

## Commands

You can also type commands. A bare command (no proposal ID) resolves to the latest in-review proposal for your vault.

| Command                                                      | Action                                                                      |
| ------------------------------------------------------------ | --------------------------------------------------------------------------- |
| `approve [PROPOSAL_ID]`                                      | Co-sign, execute, and confirm with the on-chain signature                   |
| `reject [PROPOSAL_ID]`                                       | Reject the proposal                                                         |
| `analyze [PROPOSAL_ID]`                                      | Run deep security analysis on demand                                        |
| `check pending`                                              | List in-review proposals awaiting a decision                                |
| `get status`                                                 | Screening mode, bot connection, and learned patterns                        |
| `screening on` / `screening off`                             | Toggle screening for the vault                                              |
| `update limits`                                              | Change policy limits (single-tx, daily/hourly/weekly volume, thresholds, …) |
| `list rules` / `create rule` / `update rule` / `delete rule` | Manage custom policy rules                                                  |

## Ad-Hoc Security Questions

The agent also has access to **security tools over MCP** — GoPlus address and token intelligence, Rugcheck, and sanctioned-address lookups. You can ask things like:

* "Is this address safe?"
* "Why was proposal X flagged?"
* "What's the Rugcheck score for this token?"

and get a deep report inline, even outside the context of a pending proposal.

<Note>
  Every agent request to the Sage API carries `Authorization: Bearer $AGENT_SECRET`. The agent never fabricates results — it runs the call, waits for the response, and reports the actual outcome.
</Note>