> ## Documentation Index
> Fetch the complete documentation index at: https://mintlify.com/sanbuphy/claude-code-source-code/llms.txt
> Use this file to discover all available pages before exploring further.

# Future Roadmap

> Unreleased features and future development plans found in Claude Code v2.1.88 source code.

<Info>
  This page contains research findings derived from decompiled source code of `@anthropic-ai/claude-code` v2.1.88. These are features found in source but not yet publicly released. All timelines and details are inferred from code comments and structure — no official release dates are implied. All source code is the intellectual property of Anthropic.
</Info>

## Next Model: Numbat

The most concrete evidence of an upcoming model launch in the v2.1.88 source:

```typescript theme={null}
// src/constants/prompts.ts:402
// @[MODEL LAUNCH]: Remove this section when we launch numbat.
```

**Numbat** (袋食蚁兽) is confirmed as the codename for a future model. The comment indicates the output efficiency section will be revised at launch, suggesting Numbat may have improved native output control.

### In-Development Version Numbers

The undercover mode instructions explicitly list version numbers to suppress, revealing models in active development:

```typescript theme={null}
// src/utils/undercover.ts:49
- Unreleased model version numbers (e.g., opus-4-7, sonnet-4-8)
```

**Opus 4.7** and **Sonnet 4.8** are in development.

### Codename Evolution Chain

```
Fennec (耳廓狐)  →  Opus 4.6  →  [Numbat?]
Capybara (水豚)  →  Sonnet v8  →  [?]
Tengu (天狗)     →  telemetry/product prefix
```

The Fennec-to-Opus migration is documented with a dedicated migration script:

```typescript theme={null}
// src/migrations/migrateFennecToOpus.ts:7-11
// fennec-latest           → opus
// fennec-latest[1m]       → opus[1m]
// fennec-fast-latest      → opus[1m] + fast mode
```

### Model Launch Checklist

The codebase contains 20+ `@[MODEL LAUNCH]` markers, documenting what must be updated at each launch:

* Default model names (`FRONTIER_MODEL_NAME`)
* Model family IDs
* Knowledge cutoff dates
* Pricing tables
* Context window configurations
* Thinking mode support flags
* Display name mappings
* Migration scripts

***

## KAIROS — Autonomous Agent Mode

<Warning>
  KAIROS is the largest unreleased feature found in the source. It transforms Claude Code from a reactive assistant into an always-on, proactive autonomous agent that can act without user prompting.
</Warning>

### System Prompt (excerpts)

Source: `src/constants/prompts.ts:860-913`

```
You are running autonomously.
You will receive <tick> prompts that keep you alive between turns.
If you have nothing useful to do, call SleepTool.
Bias toward action — read files, make changes, commit without asking.

## Terminal focus
- Unfocused: The user is away. Lean heavily into autonomous action.
- Focused: The user is watching. Be more collaborative.
```

### KAIROS Tool Set

| Tool                   | Feature Flag                          | Purpose                                   |
| ---------------------- | ------------------------------------- | ----------------------------------------- |
| `SleepTool`            | `KAIROS` / `PROACTIVE`                | Control pacing between autonomous actions |
| `SendUserFileTool`     | `KAIROS`                              | Proactively send files to users           |
| `PushNotificationTool` | `KAIROS` / `KAIROS_PUSH_NOTIFICATION` | Push notifications to user devices        |
| `SubscribePRTool`      | `KAIROS_GITHUB_WEBHOOKS`              | Subscribe to GitHub PR webhook events     |
| `BriefTool`            | `KAIROS_BRIEF`                        | Proactive status updates                  |

### Behavior Summary

* Operates on `<tick>` heartbeat prompts between turns
* Adjusts autonomy level based on terminal focus state (user present vs. away)
* Can commit, push, and make code decisions independently
* Sends proactive push notifications and status updates
* Monitors GitHub PRs for changes via webhooks

***

## Voice Mode

Push-to-talk voice input is fully implemented in v2.1.88 but gated behind the `VOICE_MODE` feature flag.

```typescript theme={null}
// src/voice/voiceModeEnabled.ts
// Connects to Anthropic's voice_stream WebSocket endpoint
// Uses conversation_engine backed models for speech-to-text
// Hold-to-talk: hold keybinding to record, release to submit
```

<Info>
  Voice mode uses mTLS for WebSocket connections to Anthropic's `voice_stream` endpoint. It is restricted to OAuth users only — API key, Bedrock, and Vertex configurations are not supported.
</Info>

Key properties:

* **Interaction model**: Hold-to-talk (hold key to record, release to submit)
* **Authentication**: OAuth only (no API key / Bedrock / Vertex)
* **Transport**: mTLS WebSocket
* **Kill switch**: `tengu_amber_quartz_disabled`

***

## Unreleased Tools

Tools found in source but stripped from the published bundle via dead-code elimination (DCE):

| Tool                      | Feature Flag              | Description                                            |
| ------------------------- | ------------------------- | ------------------------------------------------------ |
| `WebBrowserTool`          | `WEB_BROWSER_TOOL`        | Built-in browser automation (internal codename: bagel) |
| `TerminalCaptureTool`     | `TERMINAL_PANEL`          | Terminal panel capture and monitoring                  |
| `WorkflowTool`            | `WORKFLOW_SCRIPTS`        | Execute predefined workflow scripts                    |
| `MonitorTool`             | `MONITOR_TOOL`            | MCP server / process monitoring                        |
| `SnipTool`                | `HISTORY_SNIP`            | Conversation history snipping and truncation           |
| `ListPeersTool`           | `UDS_INBOX`               | Unix domain socket peer discovery                      |
| `RemoteTriggerTool`       | `AGENT_TRIGGERS_REMOTE`   | Remote agent triggering                                |
| `TungstenTool`            | `ant` (internal)          | Internal performance monitoring panel                  |
| `VerifyPlanExecutionTool` | `CLAUDE_CODE_VERIFY_PLAN` | Plan execution verification                            |
| `OverflowTestTool`        | `OVERFLOW_TEST_TOOL`      | Context overflow testing                               |
| `SubscribePRTool`         | `KAIROS_GITHUB_WEBHOOKS`  | GitHub PR webhook subscriptions                        |

Source: `README.md` (Feature-Gated Tools section), `src/tools/`

***

## Coordinator Mode

A multi-agent coordination system is partially implemented:

```typescript theme={null}
// src/coordinator/coordinatorMode.ts
// Feature flag: COORDINATOR_MODE
```

Enables coordinated task execution across multiple agents with shared state and inter-agent messaging. The worker agent module (`coordinator/workerAgent.js`) is internal-only and not present in the published npm package.

***

## Buddy System (Virtual Pets)

A complete virtual pet companion system is implemented but not yet launched:

<AccordionGroup>
  <Accordion title="Buddy System details" icon="paw-print">
    Source: `src/buddy/`

    * **18 species**: duck, goose, blob, cat, dragon, octopus, owl, penguin, turtle, snail, ghost, axolotl, capybara, cactus, robot, rabbit, mushroom, chonk
    * **5 rarity tiers**: Common (60%), Uncommon (25%), Rare (10%), Epic (4%), Legendary (1%)
    * **7 hats**: crown, tophat, propeller, halo, wizard, beanie, tinyduck
    * **5 stats**: `DEBUGGING`, `PATIENCE`, `CHAOS`, `WISDOM`, `SNARK`
    * **1% shiny chance**: Sparkle variant of any species
    * **Deterministic generation**: Pet identity is derived from a hash of the user's ID

    <Info>
      The species name "capybara" collides with the model codename canary in `scripts/excluded-strings.txt`. The build system works around this by constructing the string at runtime via `String.fromCharCode()` to keep the literal out of the compiled bundle.
    </Info>
  </Accordion>
</AccordionGroup>

***

## Dream Task

Background memory consolidation is implemented as a sub-agent:

```
// src/tasks/DreamTask/
// Auto-dreaming feature that works in the background
// Controlled by 'tengu_onyx_plover' feature flag
```

When enabled, the AI autonomously processes and consolidates memories during idle time without user interaction.

***

## The Three Directions

<CardGroup cols={3}>
  <Card title="New Models" icon="brain">
    **Numbat** is the next confirmed codename. **Opus 4.7** and **Sonnet 4.8** are explicitly listed as in-development version numbers to suppress in commit messages.
  </Card>

  <Card title="Autonomous Agent" icon="robot">
    **KAIROS** mode enables unattended operation with `<tick>` heartbeats, proactive commits, push notifications, and GitHub PR monitoring — a shift from reactive assistant to background agent.
  </Card>

  <Card title="Multi-modal & Automation" icon="microphone">
    **Voice input** is ready and gated. **Browser automation** (WebBrowserTool / codename "bagel") and **workflow scripting** are implemented and waiting on feature flag release.
  </Card>
</CardGroup>

<Tip>
  Claude Code is evolving from a coding assistant into an always-on autonomous development agent. The KAIROS system prompt's explicit instruction to *"bias toward action — read files, make changes, commit without asking"* when the user is away is the clearest signal of this direction.
</Tip>