Highly needed features

What variant of Codex are you using?

IDE Extension

What feature would you like to see?

Feature Requests for Codex / Chat Interface Improvements

Context & Importance:
After extensive daily use in heavy, real-world workflows, these missing features are consistently slowing down productivity and limiting effective usage. They are not edge cases—they are fundamental needs for developers and advanced users. Implementing them would significantly improve efficiency, debugging capability, and overall experience.

---

1. In-Chat Search Functionality

Description:
Add a search bar inside each conversation that allows users to quickly find specific messages, keywords, or topics without scrolling through long chat histories.

Why this matters:

• Conversations can become very long and difficult to navigate
• Improves productivity and usability
• Essential for debugging, reviewing, and referencing past outputs

Suggested Features:

• Keyword search with highlighting
• Jump-to-message navigation
• Filters (user messages / AI responses / tool outputs)

---

2. Advanced Conversation Export Options

Description:
Provide flexible export options for conversations, allowing users to choose the level of detail and format.

Export Modes:

• Compact context (current effective context only)
• Full conversation history
• Human ↔ AI messages only
• Full trace (including tool calls, system messages, links, metadata)

Why this matters:

• Enables sharing conversations with others
• Helps analyze hallucinations and reasoning issues
• Allows re-importing into AI systems for debugging or continuation

Suggested Formats:

• JSON (structured, machine-readable)
• Markdown (human-readable)
• Raw trace format (for advanced users)

---

3. Replay / Branch from Specific Responses

Description:
Allow users to replay or branch from any previous AI response or message.

Capabilities:

• “Reply from here” functionality
• Regenerate from a specific point in the conversation
• Create alternate branches without losing original thread

Why this matters:

• Helps explore different reasoning paths
• Useful for debugging incorrect outputs
• Prevents needing to restart entire conversations

---

4. Increased Sub-Agent Limit

Description:
Expand the number of supported sub-agents beyond the current limit (e.g., more than six).

Why this matters:

• Complex workflows require multiple specialized agents
• Enables more advanced orchestration and task decomposition
• Improves scalability for power users

Suggested Improvement:

• Increase limit or remove hard cap
• Allow dynamic allocation of sub-agents

---

5. Detailed Token Usage Transparency

Description:
Provide clearer and more detailed token usage metrics.

Current Issue:

• Only shows approximate usage (e.g., % or “K tokens”)
• No clear breakdown of input vs output

Requested Features:

• Separate display of:
• Input tokens
• Output tokens
• System/tool tokens
• Per-message token usage
• Visualization (e.g., graphs or breakdown bars)

Why this matters:

• Critical for debugging context limits
• Helps optimize prompt efficiency
• Improves understanding of model behavior

---

6. Per-Message Token Usage Visibility

Description:
Allow users to inspect token usage for each individual message or action in the conversation.

Suggested Implementation:

• Add a “View token usage” button on each message
• Show a breakdown per message:
• Prompt (input) tokens
• Completion (output) tokens
• Tool/system tokens (if applicable)

Optional Enhancements:

• Hover tooltip with quick token info
• Expandable panel for detailed breakdown
• Ability to aggregate token usage across selected messages

Why this matters:

• Enables fine-grained debugging of token consumption
• Helps users understand which parts of a conversation are expensive
• Useful for optimizing prompts and workflows

---

7. Configurable Sub-Agent Models and Thinking Levels

Description:
Allow users to configure the model and reasoning depth (“thinking level”) for each sub-agent individually.

Current Limitation:

• Sub-agents use a default configuration with limited user control

Requested Features:

• Select model per sub-agent (e.g., lightweight vs advanced models)
• Adjustable reasoning depth / thinking level per agent
• Presets (fast, balanced, deep reasoning)
• Override defaults when needed

Why this matters:

• Enables better cost-performance tradeoffs
• Allows specialization of agents for different tasks
• Improves control over latency and reasoning quality

---

8. Completion Notifications and External Callbacks

Description:
Add a reliable notification system that alerts users when a task, run, or response is finished.

Requested Behavior:

• Play an audible sound when execution completes
• Show a visual completion notification
• Support browser, desktop, or mobile push notifications
• Optionally trigger a webhook or URL callback when finished

Suggested Implementation:

• Completion sound toggle in settings
• Notification options for success, failure, or interruption
• Optional webhook support for advanced workflows
• Ability to send a push notification to an external app or service

Why this matters:

• Users can step away without constantly checking the screen
• Useful for long-running tasks and background workflows
• Webhook-style callbacks make automation much easier
• Improves integration with external systems and productivity tools

---

Final Note

These are not minor improvements—they are essential capabilities that directly impact real developer workflows. Many users rely on these systems heavily every day, and the absence of these features creates friction that slows down serious usage.

If there is any way to prioritize or consider implementing these, it would be deeply appreciated by the community. These changes would meaningfully improve usability, transparency, and developer efficiency. Thank you for taking the time to review them.

Additional information

_No response_