← Back to Agent Builder

Combine multiple code agents for cross-code analysis

Your key is stored locally and sent directly to OpenAI. Get an API key

➕ Add Agent Files
🧠

Knowledge Base

0 agents (0 active)
📋

No agents loaded

Upload .md agent files to build your knowledge base

⬇
🤖

Orchestrator AI

Waiting for agents...
🧠 0

Choose how the Orchestrator builds context and diagnostics for responses.

Core Orchestration (Managed by Processing Mode)

These core toggles follow the mode above and stay on in the recommended hybrid setting.

Builds a retrieval-aware prompt in the background for diagnostics only.

Uses the retrieval prompt builder for live LLM calls instead of the legacy context window.

Tracks focus episodes and summaries without persisting them.

Caps prompt size and trims context to stay within budget.

User Controls

Optional toggles you can change without affecting the mode presets.

Persists completed focus episodes into memory slices.

Shows memory slices, focus stats, and shadow prompt diagnostics.

Context Window Gauge
Healthy 0 / 0
Current prompt (estimated)
0%
Full source text (worst-case)
0%
Estimate will appear after you add code agents and a prompt.
🤖

Hello! I'm your Orchestrator AI

I have access to your knowledge base. Ask me about requirements, parameters, conflicts, or cross-references from your code agents.

Press Enter to send, Shift+Enter for new line

Analyze requirements, parameters, and conflicts across all active agents