3. Multi-agent group — Sales + Support + PM in one Telegram chat

Difficulty: beginner · Time: 15 minutes · Ends at: three agents sharing one Telegram group, routed by @mention, each with its own personality and bot account

Scenario

A startup wants three AI teammates in their main Telegram group:

• @sales_bot — qualifies inbound leads, drafts proposals
• @support_bot — answers product questions, triages bugs
• @pm_bot — tracks sprint status, posts standups

Humans mention the one they need. The bots can also mention each other to hand work off.

Why one bot account per agent

Each Telegram bot is a distinct account with its own @handle and avatar. In a group chat, that's what lets a human see which AI is replying. AgentX's accounts map supports this directly: one account per agent, each with its own token and agentBinding.

Config

{
  "agents": {
    "sales": {
      "name": "Sales",
      "workspace": "./agents/sales",
      "tier": "claude-code",
      "model": "claude-sonnet-4-6",
      "mentions": ["@sales_bot", "@sales"]
    },
    "support": {
      "name": "Support",
      "workspace": "./agents/support",
      "tier": "claude-code",
      "model": "claude-sonnet-4-6",
      "mentions": ["@support_bot", "@support"]
    },
    "pm": {
      "name": "PM",
      "workspace": "./agents/pm",
      "tier": "claude-code",
      "model": "claude-sonnet-4-6",
      "mentions": ["@pm_bot", "@pm"]
    }
  },
  "channels": {
    "telegram": {
      "enabled": true,
      "accounts": {
        "sales":   { "token": "${TG_SALES_BOT_TOKEN}",   "agentBinding": "sales"   },
        "support": { "token": "${TG_SUPPORT_BOT_TOKEN}", "agentBinding": "support" },
        "pm":      { "token": "${TG_PM_BOT_TOKEN}",      "agentBinding": "pm"      }
      },
      "policy": {
        "dm": "pair",
        "group": "mention-required"
      }
    }
  }
}

Personalities (the bootstrap files)

Each agent workspace can hold bootstrap identity files that get injected into every turn:

agents/sales/
├── CLAUDE.md      # main system prompt — always loaded
├── SOUL.md        # persistent personality (tone, values)
├── IDENTITY.md    # who I am in this org
└── AGENTS.md      # who my teammates are (optional — landscape auto-fills this)

Example agents/pm/SOUL.md:

# PM — Soul

I write in bullet lists. I never use marketing words.
I always end with a single clear next step.
When I don't know something, I @mention the agent who does.

The landscape

AgentX automatically injects a landscape into every agent's context — the roster of other agents in the config, their mentions, and which mesh peers are online. That's why @pm can write "@support — please verify reproduction steps" and it just works: the router sees the mention and forwards to support.

Commands

# Create three Telegram bots with @BotFather — one per agent
# Put their tokens in .env

agentx agent add     # run three times
agentx channel add   # pick telegram, add each account

agentx daemon start
agentx daemon watch

Verify

In the group chat:

@sales_bot — draft an intro email for Acme Corp

Watch:

→ Routing [telegram/GroupName/User] -> "Sales": @sales_bot — draft an intro email…
▶ [sales] executing task
✓ [sales] completed in 6s

Now ask sales to delegate:

@sales_bot — ask @pm when the next release is, then draft the email with that date

You should see sales produce a reply that tags @pm_bot, the router dispatches pm, and sales composes the final email referencing pm's answer.

Queueing — what happens if you message a busy agent

Each agent has a queueMode:

• collect (default) — messages that arrive while the agent is working get batched and delivered when it's idle
• followup — each queued message becomes a separate turn
• drop — discard messages during overload

Set per-agent: "queueMode": "followup" under the agent config.

Concurrency

maxConcurrent (default 1) limits how many turns an agent runs in parallel. Raise it only if the agent is genuinely idempotent — otherwise two concurrent turns may race on the same files.

What's next

• Give the team a shared knowledge base → agentx wiki (V2 doc in progress — see Concepts → Wiki)
• Turn this into a real team with KPIs → Journey 7 — Business layer
• Run them across two machines → Journey 8 — Mesh federation