Skip to content

Latest commit

 

History

History
57 lines (44 loc) · 2.36 KB

File metadata and controls

57 lines (44 loc) · 2.36 KB

Features

A detailed tour of what AgentForge can do. Everything runs client-side from static files (index.html + assets/), with no build step or backend.

💬 Chat

Talk to any of the 247 specialists. Responses stream token-by-token. Each agent carries a deep, role-specific system prompt — not a generic assistant persona.

⛓️ Agent Pipeline

Chain 2–6 specialists into a sequence. Each agent receives all previous outputs as context and builds on them. Example: Product Manager → Frontend Developer → Brand Guardian → Growth Hacker produces a coherent launch plan in one run.

✨ Auto-Build Team

Describe a goal in plain language. An orchestrator reads the entire roster and selects the optimal agents in the right order, so you don't need to know which specialists to pick.

⚔️ Agent Debate

Choose two specialists. They argue in rounds — a Proposer builds a position, a Critic challenges it, the Proposer refines — until a battle-tested answer emerges.

📊 Benchmarker

Run the same prompt across several agents. An impartial judge model scores each response on relevance, depth, and correctness (not length), so you can see which specialist actually performs best for a task.

▶️ Code Runner

Every code block the AI produces gets a Run button. Execution happens via the free Piston API across 70+ languages (Python, JavaScript, Go, Rust, Java, C++, and more) — no local setup.

🌐 Web Search

Optionally give agents real-time information via the Brave Search API. Bring your own free key and add it in Settings.

🎤 Voice Input

Tap the mic and speak; your words fill the message box. Uses the built-in Web Speech API in Chrome and Safari — no extra install.

📎 File Attachment

Attach code files, CSVs, documents, or notes. The agent reads them as context. Fully client-side via FileReader — nothing is uploaded.

🧠 Memory

Save any chat or pipeline output to persistent memory (localStorage). Review, copy, or delete entries anytime.

⬇️ Export

Download any chat, pipeline, debate, or benchmark report as a .md file.

🔄 Smart Rate-Limit Handling

On Groq's free tier, if one model hits its per-minute token limit, AgentForge waits the required time and rotates to another model in a 4-model pool — so pipelines finish without manual retries. See ARCHITECTURE.md.