<p>
  <img src="banner.png" alt="pi-intercom" width="1100">
</p>

# Selesai Intercom

Direct 1:1 messaging between Selesai sessions on the same machine. Send context, findings, or requests from one session to another — whether you're driving the conversation or letting agents coordinate.

```text
User flow: press Alt+M or run /intercom to pick a session and send a message
```

## Why

Sometimes you're running multiple Selesai sessions — one researching, one executing, one reviewing. pi-intercom lets you:

- **User-driven orchestration** — Send context or findings from your research session to your execution session
- **Agent collaboration** — An agent can reach out to another session when it needs help or wants to share results
- **Session awareness** — See what other Selesai sessions are running and their current status

Unlike pi-messenger (a shared chat room for multi-agent swarms), pi-intercom is for targeted 1:1 communication where you pick the recipient.

pi-intercom also integrates well with [pi-subagents](https://github.com/nicobailon/pi-subagents): delegated child agents get a child-only `contact_supervisor` tool when `pi-subagents` supplies bridge metadata. Use `reason: "need_decision"` for blocking clarification, `reason: "interview_request"` for multiple structured supervisor answers, and `reason: "progress_update"` for meaningful plan-changing updates. Normal sessions only see the regular `intercom` tool.

## In One Minute

Each Selesai session that has `pi-intercom` loaded and enabled connects to a tiny local broker over a local IPC transport. The broker keeps track of connected sessions and routes direct messages to the one you target by name or session ID. The extension gives you both a tool (`intercom`) and a small overlay UI (`/intercom` or `Alt+M`). Incoming messages are rendered inline inside the recipient session, can trigger a turn immediately by default, and are also stored in Selesai session history as extension entries. If you want a stricter local trust posture, `inboundTrigger` can reduce or disable auto-triggering.

## Install

```bash
selesai install npm:pi-intercom
```

Then restart Selesai. The extension auto-connects to the broker on startup and registers the bundled `pi-intercom` skill for common coordination patterns.

**Recommended:** Add this snippet to your project's `AGENTS.md` to help agents understand when to coordinate across sessions:

```xml
<pi-intercom>
Coordinate with other local Selesai sessions on related codebases. Use `/skill:pi-intercom` for patterns.

**When:** Same codebase (parallel work), reference codebase (consulting patterns), related repos (shared libraries).

**Not when:** Unrelated codebases, trivial questions, or when you can proceed independently.

**Principle:** Prefer `send` for notifications; `ask` only when blocked waiting for input.
</pi-intercom>
```

A session becomes intercom-connected when all of these are true:
- the `pi-intercom` extension is installed and loaded in that session
- `enabled` is not set to `false` in the intercom config file, which defaults to `~/.selesai/agent/intercom/config.json`
- the session has started or reloaded after the extension was installed
- the local broker is running or can be auto-started

The session list only shows intercom-connected sessions, not every open Selesai process on the machine.

If a session is unnamed, pi-intercom exposes a collision-resistant runtime-only fallback alias like `subagent-chat-1a2b3c4d-5e6f-7a8b` so other connected sessions can target it. That alias is not persisted as the Selesai session title or treated as a reconnect identity, so `selesai --resume` can keep showing the transcript snippet without allowing a different unnamed process to inherit queued mail.

## Quick Start

### From the Keyboard

Press **Alt+M** or type `/intercom` to open the session list overlay:

1. **Select a session** — Use arrow keys to pick a target session
2. **Compose message** — Write your message in the compose overlay
3. **Send** — Press Enter to send, Escape to cancel

### From the Agent

The agent can list sessions and send messages using the `intercom` tool. Tool calls and results render as compact transcript rows so send/ask/reply flows are easy to scan. Use `/intercom-id` to insert a handoff snippet for the current session's stable intercom target into the editor. For common patterns like planner-worker delegation, the bundled `pi-intercom` skill provides copy-paste ready examples:

```typescript
// List active sessions
intercom({ action: "list" })
// → **Current session:**
// → • executor (20d43841) — ~/projects/api (claude-sonnet-4 · 42% ctx) [self, idle]
// → **Other sessions:**
// → • research (6332faab) — ~/projects/api (claude-sonnet-4) [same cwd, thinking]

// List only peers in the same working directory
intercom({ action: "list-cwd" })

// Send a message
intercom({ action: "send", to: "research", message: "Check if UserService.validate() handles null" })
// → Message sent to research

// Send to a visible peer session in another codebase, opening a Herdr project pane if needed
intercom({
  action: "send",
  cwd: "/Users/me/projects/billing",
  openProjectPaneIfMissing: true,
  message: "Let's discuss the billing retry design in this repo."
})
// → Opened Herdr project pane pane-... for /Users/me/projects/billing and sent message to subagent-chat-...

// Check connection status
intercom({ action: "status" })
// → Connected: Yes, Session ID: abc123, Active sessions: 3

// Send with attachments (code snippets, files, or context)
intercom({
  action: "send",
  to: "worker",
  message: "Here's the fix:",
  attachments: [{
    type: "snippet",
    name: "auth.ts",
    language: "typescript",
    content: "function validate(user: User) { ... }"
  }]
})
```

### Receiving Messages

When a message arrives, it appears inline in your chat with the sender's info and a reply hint:

```
**From research** (~/projects/api)

To reply, use the intercom tool: intercom({ action: "reply", message: "..." })

Found the issue — UserService.validate() doesn't check for null input.
See auth.ts:142-156.
```

The reply hint (enabled by default) points to `intercom({ action: "reply", ... })`, so recipients do not need raw sender or `replyTo` IDs. Idle recipients get a new turn immediately; busy interactive recipients receive the message through Selesai's steering queue at the next safe model boundary without aborting the active turn. Attachment content is included in the agent-visible body, and messages are rendered inline and stored in Selesai session history.

## Workflow: Planner-Worker Coordination

The most natural use of pi-intercom is splitting a task between two sessions — one holds the big picture, the other does the hands-on work. When the worker hits an ambiguity ("should I optimize for readability or performance here?"), they ask without losing context.

### Setup

Open two terminals and start pi in each. Name them so they can find each other:

```
# Terminal 1                    # Terminal 2
/name planner                   /name worker
```

Verify they see each other from either session:

```typescript
intercom({ action: "list" })
// → • worker — ~/projects/api (claude-sonnet-4) [idle]
```

### The Conversation

Here's how a typical exchange looks. The planner delegates with `send` (fire-and-forget). The worker uses `ask` for anything that needs a response — questions, discoveries, completion reports. `ask` sends the message and blocks until the planner replies, so the worker gets the answer as a tool result and continues in the same turn.

**Planner sends a task:**
```typescript
intercom({
  action: "send",
  to: "worker",
  message: "Task-3: Add retry logic to API client. Key files: src/api/client.ts, src/api/types.ts. Ask if anything's unclear."
})
```

**Worker hits an ambiguity — asks and waits:**
```typescript
intercom({
  action: "ask",
  to: "planner",
  message: "Should retry apply to all endpoints or just idempotent ones? Also, max retry count and backoff strategy?"
})
// → Reply from planner: Only GET/PUT/DELETE — never POST. Max 3 retries, exponential backoff starting at 100ms.
// Worker continues implementing with the answer, same turn, full context.
```

**Worker finds something unexpected — escalates and waits:**
```typescript
intercom({
  action: "ask",
  to: "planner",
  message: "Found: fetchWithTimeout swallows network errors. Fixing this changes the error shape. OK to proceed?"
})
// → Reply from planner: Yes, surface the error types. The current behavior is a bug.
```

**Worker reports completion:**
```typescript
intercom({
  action: "ask",
  to: "planner",
  message: "Task-3 done. Added RetryPolicy type, applied to GET/PUT/DELETE, surfaced NetworkError, 4 tests passing."
})
// → Reply from planner: Looks good. Move on to task-4.
```

### Communication Patterns

| Pattern | Action | Why |
|---------|--------|-----|
| **Task Delegation** | Planner uses `send` | Fire-and-forget. Planner doesn't need to wait for an ack. |
| **Clarification Request** | Worker uses `ask` | Worker needs the answer to proceed. Blocks until reply. |
| **Discovery Escalation** | Worker uses `ask` | Worker needs approval before changing course. |
| **Completion Report** | Worker uses `ask` | Planner might have follow-up instructions or the next task. |

### Reply Hints

When `replyHint` is enabled (the default), incoming messages include the exact `intercom()` call to respond:

```
**From planner** (~/projects/api)

To reply, use the intercom tool: intercom({ action: "reply", message: "..." })

Only GET/PUT/DELETE — never POST. Max 3 retries with exponential backoff starting at 100ms.
```

This matters because the agent receiving the message doesn't need to reconstruct raw `to` and `replyTo` IDs — the hint is right there. Combined with immediate idle triggering and safe busy-turn steering, it enables real back-and-forth conversation without aborting work in progress or delaying messages until they become stale. If the reply happens later instead of in the triggered turn, `intercom({ action: "reply" })` falls back to the single unresolved inbound ask, and `intercom({ action: "pending" })` shows who is still waiting.

### `send` vs `ask`

`send` is fire-and-forget — the tool returns immediately after delivery. When the destination has exactly one pending inbound ask, `send` infers that it is the answer, attaches the ask's `replyTo`, and reports `Reply sent to <target> (inferred from pending ask)`. With zero or multiple matching asks, it remains an ordinary unthreaded send. An inferred answer still uses the `confirmSend` dialog when configured; only a caller-supplied `replyTo` skips confirmation.

`ask` requires a currently connected recipient, then blocks until it responds (10-minute timeout by default; set `PI_INTERCOM_ASK_TIMEOUT_MS` to a positive millisecond value to change it). If the target is disconnected, `ask` fails immediately; use `send` when queued, non-blocking mailbox delivery is appropriate. The reply comes back as the tool result, so the agent continues in the same turn with full context. No confirmation dialog — if you're asking and waiting, the intent is clear.

`reply` is receiver-side sugar for replying to an inbound ask. In the turn triggered by an incoming intercom ask, `intercom({ action: "reply", message: "..." })` targets that exact sender and message automatically. If you reply later, it falls back to the single unresolved inbound ask. If multiple asks are pending, use `intercom({ action: "pending" })` to inspect them and then call `reply` with `to` to disambiguate.

The broker keeps a bounded in-memory mailbox for recently disconnected explicitly named sessions. If a lightweight CLI sender asks a long-running session something and exits before the answer, the later `reply` is accepted into that mailbox instead of failing with `Session not found`; a process that reconnects with the same explicit name and directory receives the queued reply. Runtime-only unnamed-session aliases never transfer mailbox ownership, and routing never remaps mail back to its sender. This is per-broker runtime state, not durable storage across broker restarts.

Incoming messages carry diagnostic metadata end to end: stable message ID, sender sequence, sender timestamp, broker receive/delivery timestamps, receiver receive timestamp, and injection timestamp. Connected interactive receivers emit `receiver_received`, `acknowledged`, and `injected` as they hand messages to Pi; duplicate IDs are acknowledged but injected at most once per receiving session. Broker mailbox delivery for temporarily disconnected targets can still report queued delivery. If an `ask` times out, the timeout names the message ID and last known delivery state. Timeout is not cancellation: an injected or broker-queued message may remain actionable unless an explicit cancellation path says otherwise.

Cancellation is explicit: call `intercom({ action: "cancel", messageId })` to request cancellation of a message you originally sent. Connected interactive messages are injected immediately, so the receiver normally reports `cancellation_requested` rather than pretending it removed work from a private queue. Supersede is also explicit: pass `supersedes: "old-message-id"` on a new `send` or `ask`. The broker only allows same sender → same receiver supersedes, marks the old message `superseded`, and sends the replacement with a new ID; an already-steered old message may still be processed. Retries are never automatic; a retry should be a new authored message, optionally linked with `retryOf`.

The planner typically uses `send`. If you prefer manual approval for outgoing non-reply messages, turn on `confirmSend: true`. The worker uses `ask` for everything (no confirmation needed, gets answers inline), so it can operate autonomously either way.

## Workflow: Subagent-to-Supervisor Escalation

This workflow requires [`pi-subagents`](https://github.com/nicobailon/pi-subagents) to be installed and to supply child bridge metadata. When `pi-subagents` spawns a delegated child with that metadata, the child session gets a subagent-only `contact_supervisor` tool in addition to the regular `intercom` tool. Normal sessions never see `contact_supervisor`.

### When the Tool Appears

`contact_supervisor` only registers when `pi-subagents` sets all of these environment variables:

- `SELESAI_SUBAGENT_ORCHESTRATOR_TARGET` — the supervisor session name or ID
- `SELESAI_SUBAGENT_RUN_ID` — the run identifier
- `SELESAI_SUBAGENT_CHILD_AGENT` — the agent type
- `SELESAI_SUBAGENT_CHILD_INDEX` — the child index within the run

If any are missing, the session falls back to the regular `intercom` tool.

### Three Reasons

| Reason | Behavior | Use When |
|--------|----------|----------|
| `need_decision` | Sends an ask and blocks until the supervisor replies (10-minute timeout by default; configurable with `PI_INTERCOM_ASK_TIMEOUT_MS`) | The subagent is blocked, uncertain, needs approval, or faces a product/API/scope decision |
| `interview_request` | Sends structured questions and blocks until the supervisor replies | The subagent needs multiple machine-readable answers from the supervisor in one exchange |
| `progress_update` | Fire-and-forget update to the supervisor | Meaningful progress or unexpected discoveries that change the plan |

Do not use `contact_supervisor` for routine completion handoffs. Return the final subagent result normally through `pi-subagents`.

A child subagent can still use the regular `intercom` tool to coordinate with an explicit peer session. Use `to` alone for any live peer on the machine, `cwd` alone for the sole live peer in another codebase, or `to` plus `cwd` when the directory is a safety guard. Use `contact_supervisor` instead when the answer changes the task contract, needs owner approval, or would require opening a new visible project pane.

For bounded work in another codebase, prefer `pi-subagents` with an explicit `cwd`. Intercom project panes are for durable visible peer conversations, not ordinary delegated work.

### Example: Blocked Subagent Asks for Guidance

```typescript
contact_supervisor({
  reason: "need_decision",
  message: "The auth service returns 403 instead of 401 for expired tokens. Should I treat 403 as a re-auth trigger or a hard failure?"
})
// → Reply from supervisor: Treat 403 as re-auth trigger. Update the token refresh logic.
```

### Example: Structured Supervisor Interview

```typescript
contact_supervisor({
  reason: "interview_request",
  message: "Please answer these before I continue the migration.",
  interview: {
    title: "API migration choices",
    questions: [
      { id: "api", type: "single", question: "Which API should I target?", options: ["Stable API", "Experimental API"] },
      { id: "constraints", type: "text", question: "What constraints should I preserve?" }
    ]
  }
})
// → Reply from supervisor: { "responses": [{ "id": "api", "value": "Stable API" }, ...] }
```

### Example: Progress Update

```typescript
contact_supervisor({
  reason: "progress_update",
  message: "Discovered the bug is in the retry wrapper, not the API client. Fixing the wrapper will also close issue #42."
})
// → Progress update sent to supervisor planner
```

### What the Supervisor Sees

The supervisor receives a formatted message with run metadata:

```
**From subagent-worker-78f659a3-1**

Subagent needs a supervisor decision.
Run: 78f659a3
Agent: worker
Child index: 0

Which API should I use?
```

Reply hints work the same as regular `intercom` ask/reply flows. The supervisor can reply with `intercom({ action: "reply", message: "..." })` and the subagent receives the answer as the tool result.

For `interview_request`, the supervisor message includes the structured questions plus a fenced JSON answer example using this stable shape:

```json
{
  "responses": [
    { "id": "api", "value": "Stable API" },
    { "id": "constraints", "value": "Keep the public error shape unchanged." }
  ]
}
```

The supervisor can reply with plain JSON or a fenced `json` block. If the reply matches the `{ "responses": [...] }` shape and references valid question ids/options, the child tool result includes it in `details.structuredReply` while still showing the raw reply text.

## Tool Reference

### intercom

| Parameter | Type | Description |
|-----------|------|-------------|
| `action` | string | `"list"`, `"list-cwd"`, `"send"`, `"ask"`, `"reply"`, `"pending"`, `"status"`, or `"cancel"` |
| `to` | string | Target session name or ID. Without `cwd`, send/ask resolve it globally. With `cwd`, send/ask require the target to be in that directory. Also disambiguates reply. |
| `message` | string | Message text (for send/ask/reply) |
| `attachments` | array | Optional `file`, `snippet`, or `context` attachments |
| `replyTo` | string | Optional message ID for threading or replying to an `ask` |
| `messageId` | string | Optional explicit message ID for send/ask, or required message ID for `cancel` |
| `supersedes` | string | Optional previous message ID that this send/ask explicitly replaces |
| `retryOf` | string | Optional previous message ID that this send/ask explicitly retries |
| `cwd` | string | Working directory filter for `list-cwd`. For send/ask, scopes target lookup to that directory; without `to`, selects the sole live peer there. |
| `openProjectPaneIfMissing` | boolean | For `send`/`ask` with `cwd`, open a visible Herdr project pane and launch Selesai when no matching live session exists |
| `focus` | boolean | For `openProjectPaneIfMissing`, focus the new Herdr pane. Defaults to true |

### contact_supervisor

Only registered in sessions where `pi-subagents` supplied the required child bridge metadata. Contacts the supervisor session that delegated the current task.

| Parameter | Type | Description |
|-----------|------|-------------|
| `reason` | string | `"need_decision"` (blocking), `"interview_request"` (blocking structured questions), or `"progress_update"` (fire-and-forget) |
| `message` | string | The decision request, optional interview note, or progress update |
| `interview` | object | Required for `interview_request`: `{ title?, description?, questions: [...] }` |

**`need_decision`** — Sends a formatted ask to the supervisor and blocks until it replies (10-minute timeout by default; configurable with `PI_INTERCOM_ASK_TIMEOUT_MS`). The reply comes back as the tool result. Includes run metadata in the message so the supervisor knows which subagent is asking.

**`interview_request`** — Sends a formatted, agent-readable interview to the supervisor and blocks until it replies. Questions use a local pi-interview-like shape: `{ id, type, question, options?, context? }` where `type` is `single`, `multi`, `text`, `image`, or `info`. `info` questions are context-only and do not need responses. The supervisor reply should be JSON with `{ "responses": [{ "id": "...", "value": ... }] }`. Parsed JSON replies are returned in `details.structuredReply`.

**`progress_update`** — Sends a non-blocking update to the supervisor. Returns immediately after delivery. Use only for meaningful progress or unexpected discoveries that change the plan.

### intercom actions

**`list`** — Returns the current session plus other active intercom-connected sessions with name, short ID, working directory, model, and live status. Status is derived automatically from Selesai lifecycle events: `idle`, `thinking`, or `tool:<name>`.

**`send`** — Sends a message to the specified session and returns immediately after delivery. If the destination has exactly one pending inbound ask, `send` infers the message is its answer and returns `Reply sent to <target> (inferred from pending ask)`; zero or multiple matches remain unthreaded sends. Set `confirmSend: true` to confirm ordinary and inferred sends. A caller-supplied `replyTo` skips confirmation. `to` alone resolves globally across all live sessions. `cwd` alone targets the sole live peer in that directory. `to` plus `cwd` requires that peer to be in the directory. With `openProjectPaneIfMissing: true`, pi-intercom opens a visible Herdr project pane, starts Selesai there, waits for that session to register, then delivers the message through normal intercom routing.

**`ask`** — Requires a currently connected recipient, sends a message, and waits for the recipient to reply (10-minute timeout by default; configurable with `PI_INTERCOM_ASK_TIMEOUT_MS`). A disconnected target fails immediately rather than queueing a blocking request. The reply is returned as the tool result. No confirmation dialog. Only one pending `ask` is allowed per session at a time. Use this when the agent needs the answer to continue working. The same `to`, `cwd`, and `openProjectPaneIfMissing` targeting rules apply.

**`reply`** — Replies to the current intercom-triggered message if there is one. Otherwise it falls back to the single unresolved inbound ask. If multiple asks are pending, pass `to` or inspect them with `pending` first. Under the hood this is still a normal `send` with the exact `replyTo` value.

**`pending`** — Lists unresolved inbound asks with sender, message ID, elapsed time, and a short preview. Useful when replying after the original triggered turn.

**`cancel`** — Requests cancellation of a message previously sent by the current session. Queued messages are removed before injection; already-injected messages receive a visible cancellation request.

**`status`** — Shows connection status, session ID, and total count of active sessions (including the current session).

## Keyboard Shortcuts

| Key | Action |
|-----|--------|
| Alt+M | Open session list overlay |
| ↑/↓ | Navigate session list |
| Enter | Select session / Send message |
| Escape | Cancel / Close overlay |

## Config

Create `~/.selesai/agent/intercom/config.json`:

```json
{
  "brokerCommand": "npx",
  "brokerArgs": ["--no-install", "tsx"],
  "confirmSend": false,
  "inboundTrigger": "always",
  "enabled": true,
  "replyHint": true,
  "status": "researching"
}
```

| Setting | Default | Description |
|---------|---------|-------------|
| `brokerCommand` | `"npx"` | Advanced trusted override for the broker executable. The default value is hardened internally to launch the resolved bundled `tsx` CLI through the current Node executable instead of resolving `npx` through `PATH`. |
| `brokerArgs` | `["--no-install", "tsx"]` | Advanced trusted arguments passed to custom `brokerCommand` before the broker script path |
| `confirmSend` | false | Show a confirmation dialog before ordinary or inferred sends from an interactive session with UI; caller-supplied `replyTo` skips it |
| `inboundTrigger` | `"always"` | Auto-trigger policy for inbound broker messages: `"always"`, `"replies"`, or `"never"`. Local in-process subagent relay events still trigger the addressed session. |
| `enabled` | true | Enable/disable intercom entirely |
| `replyHint` | true | Include reply instruction in incoming messages |
| `status` | — | Optional custom status suffix shown after the automatic lifecycle status, for example `thinking · researching` |

If `config.json` cannot be parsed or contains an invalid value, pi-intercom fails to load the config and throws an error, so the misconfiguration is surfaced immediately instead of silently changing behavior.

Custom broker commands are trusted local configuration: anyone who can edit this config can choose the executable used for future broker auto-spawns. For example, if you have Bun installed and want it to start the broker directly, use:

```json
{
  "brokerCommand": "bun",
  "brokerArgs": []
}
```

Pi-intercom publishes live session status automatically. Sessions register as `idle`, switch to `thinking` while the agent is running, show `tool:<name>` during tool execution, and return to `idle` on agent completion. If `status` is set in config, it is appended as context instead of replacing the lifecycle status.

By default, runtime state and config live under `~/.selesai/agent/intercom`. If Selesai is launched with `SELESAI_CODING_AGENT_DIR`, pi-intercom uses `$SELESAI_CODING_AGENT_DIR/intercom` instead, including `config.json`, broker PID/lock files, sockets, and launcher state.

## Extension channels

Other Selesai extensions can use intercom's broker for bounded, non-conversational coordination. Extension-channel traffic never calls `pi.sendMessage()`, never enters a session transcript, and never starts an agent turn.

Register during `session_start` so intercom includes the capability in its deferred broker registration:

```typescript
import type { ExtensionAPI } from "@selesai/code";
import {
  INTERCOM_EXTENSION_REGISTER_EVENT,
  type IntercomExtensionChannel,
} from "pi-intercom/extension-api.ts";

export default function (pi: ExtensionAPI) {
  let channel: IntercomExtensionChannel | undefined;

  pi.on("session_start", () => {
    pi.events.emit(INTERCOM_EXTENSION_REGISTER_EVENT, {
      namespace: "example/v1",
      ownerEligible: true,
      onReady: (value: IntercomExtensionChannel) => { channel = value; },
      onEvent: (event: unknown) => { /* owner, state, peer, or payload event */ },
    });
  });
}
```

The broker:

- advertises `extension-bus-v1` through feature negotiation
- routes payloads only to sessions advertising the same namespace
- elects one owner per namespace and changes its epoch after socket replacement
- rejects stale owner-only writes
- stores at most 64 KiB of opaque, revisioned state per namespace

`channel.publish()` accepts payloads up to 16 KiB. A `capable` broadcast includes the sender, so consumers must not blindly republish messages they receive. `channel.commitState()` uses compare-and-swap against the last observed revision. Capabilities registered after the broker connection is established are synchronized without reconnecting. Clients connected to an older broker see the channel as unsupported and do not send extension operations.

## How It Works

```mermaid
graph TB
    subgraph A["Selesai Session A"]
        A1[Intercom Client]
        A2[intercom tool]
        A3[UI overlays]
    end

    subgraph Broker["Intercom Broker"]
        B1[Session Registry]
        B2[Message Router]
    end

    subgraph B["Selesai Session B"]
        B3[Intercom Client]
        B4[intercom tool]
        B5[UI overlays]
    end

    A1 <-->|Local Socket/Pipe| B1
    B1 --- B2
    B2 <-->|Local Socket/Pipe| B3
```

The broker is a standalone TypeScript process that manages session registration and message routing. It auto-spawns when the first intercom-enabled session needs it and exits after 5 seconds when the last connected session disconnects. Clients now reconnect automatically if the broker disappears and later comes back.

**Liveness heartbeat.** A client whose broker is killed without a clean shutdown (SIGKILL, crash, or host loss) is left on a half-open socket: the OS never delivers a `close` event, so the client cannot tell it is alone and silently drops out of the roster forever. To close that gap, each registered client runs a liveness heartbeat that round-trips a lightweight `list` request and tears down the socket if the broker does not respond within the timeout, letting the existing `disconnected` → reconnect path fire. The interval defaults to 30s and the probe timeout to 5s; override them with `PI_INTERCOM_LIVENESS_INTERVAL_MS` and `PI_INTERCOM_LIVENESS_TIMEOUT_MS` (the timeout is clamped to the interval).

Messages use length-prefixed JSON over a local socket/pipe transport (4-byte length + JSON payload) to handle fragmentation properly. The protocol includes request correlation for session listing, explicit delivery failures, validation for malformed or out-of-order messages, a frame-size cap, per-connection local rate limiting, and no-op presence coalescing.

Session IDs are the trusted addressing key. Duplicate names remain allowed for same-user workflows, but sends to ambiguous names fail and users should target the stable session ID shown by `list`/`status` in trust-sensitive flows. Mail queued for a disconnected session is redelivered to a session that reconnects under the same session ID, or to a session that matches both its explicit name and its directory, so a same-named session in a different project never inherits another project's queued messages. Runtime-only `subagent-chat-...` aliases are excluded from name-based mailbox reconnection, and a disconnected mailbox is never remapped to the sender. Set `PI_INTERCOM_STABLE_ID` or `stableId` in `config.json` to pin a session's intercom ID across full process relaunches; `config.json` is machine-global, so a fixed `stableId` there applies to every session on the machine and the newest registration takes over that identity. The broker owns local trust metadata such as `trustedLocal`; `peerUid` is reserved for runtimes that can expose real peer credentials and is left unset otherwise. Client-supplied cwd/model/pid/status are display metadata, not authentication.

Async extension work (startup, inbound flushes, reconnects, overlays, and relays) no-ops if the session shuts down or reloads before it settles.

Runtime files live at `~/.selesai/agent/intercom/` by default, or `$SELESAI_CODING_AGENT_DIR/intercom/` when `SELESAI_CODING_AGENT_DIR` is set:
- `broker.sock` — Unix domain socket for communication (macOS/Linux only; Windows uses a named pipe instead)
- `broker-launch.vbs` — Windows helper script used to launch the broker without a console window
- `broker.pid` — Broker process ID
- `broker.spawn.lock` — Auto-spawn lock file
- `broker.port.json` — Dynamic localhost TCP endpoint, only when Windows TCP transport is explicitly enabled
- `config.json` — User configuration

Supported `config.json` keys include `stableId` for restart-stable addressing, `status` for a custom status suffix, `inboundTrigger` (`always`, `replies`, or `never`), `replyHint`, `confirmSend`, and advanced broker launch overrides.

## Design Decisions

**Local IPC instead of TCP.** Same-machine only by design. `pi-intercom` uses Unix sockets on macOS/Linux and a named pipe on Windows, which keeps setup simple and avoids port management. Windows TCP is available only as an explicit escape hatch with `PI_INTERCOM_TRANSPORT=tcp` (or `PI_INTERCOM_TCP=1`) for environments where named pipes are blocked. In that mode the broker binds a dynamic `127.0.0.1` port, records the endpoint plus a local secret under the intercom state dir, and requires that secret before health or registration succeeds. Health replies do not echo the secret, so a random localhost process cannot discover it through the broker protocol.

**Auto-spawn with file lock.** The broker starts on first connection and exits after 5 seconds idle. There is no daemon to manage. A spawn lock file, keyed by PID and timestamp, prevents duplicate brokers when multiple sessions start at once.

**`ask` stays client-side.** The broker still routes plain messages; it does not have a special request/response mode for `ask`. The client waits for a matching reply before it triggers a new turn, then returns that reply as the tool result. Reply hints make that flow practical by showing the recipient the exact `send` call to use. Separately, `list` / `sessions` now carry a `requestId` so a delayed session-list reply cannot be mistaken for a newer one.

## pi-intercom vs pi-messenger

| Aspect | pi-intercom | pi-messenger |
|--------|-------------|--------------|
| **Model** | Direct 1:1 messaging | Shared chat room |
| **Primary use** | User orchestrating sessions | Autonomous agent coordination |
| **Discovery** | Broker-based (real-time) | File-based registry |
| **Messages** | Private, session-to-session | Broadcast to all agents |
| **Persistence** | In Selesai session history | Shared coordination files |

Use pi-messenger for multi-agent swarms working on a shared task. Use pi-intercom when you want to manually coordinate your own sessions or have one agent reach out to another specific session.

## File Structure

```
~/.selesai/agent/extensions/pi-intercom/
├── package.json
├── index.ts              # Extension entry point
├── types.ts              # SessionInfo, Message, protocol types
├── config.ts             # Config loading
├── project-agent.ts      # Herdr project-pane launch and cwd target resolution
├── broker/
│   ├── broker.ts         # Broker process
│   ├── client.ts         # IntercomClient class
│   ├── framing.ts        # Length-prefixed JSON protocol
│   ├── paths.ts          # Platform-specific socket/pipe paths
│   ├── spawn.ts          # Auto-spawn logic with lock file
│   ├── spawn.test.ts     # Broker spawn tests
│   └── paths.test.ts     # Path resolution tests
├── ui/
│   ├── session-list.ts   # Session selection overlay
│   ├── compose.ts        # Message composition overlay
│   └── inline-message.ts # Received message display
└── skills/
    └── pi-intercom/
        └── SKILL.md      # Bundled skill for common patterns
```

## Limitations

- **Same machine only** — Uses local sockets/pipes, no network support
- **No dedicated intercom log** — Messages are kept in Selesai session history, but there is no separate intercom transcript or inbox
- **No attachments UI** — `file`, `snippet`, and `context` attachments are supported in the protocol, but not in the compose overlay
- **Only connected sessions appear** — The list shows Selesai sessions that have loaded `pi-intercom` and successfully registered with the broker, not every open Selesai process on the machine
- **Broker lifecycle** — The broker auto-spawns on first use and exits when idle; sessions reconnect automatically if the broker restarts