Find the Best AI Agent Behavior
— Automatically

Run live experiments, compare variants, and let your agents learn which strategy performs best — in real-time, without manual testing.

Get Started →
Works with:
Claude Code
ChatGPT MCP
LangChain

The Problem: Your AI Agent Can't Learn

Most AI agents are stuck with fixed prompts and static configurations. When you change a prompt, you're guessing whether it's better. There's no feedback loop, no learning, no adaptation.

◆
Converges Fast
Thompson Sampling finds the best variant in 50-150 evaluations. No weeks of manual A/B testing — learn what works in hours.
▸
Lightning Fast Selection
Sub-10ms variant selection means your agent never waits. Make decisions in real-time without sacrificing intelligence.
◉
Privacy First
Your experiments stay yours. Automatic 7-day expiration. Isolated by experiment ID. No cross-experiment data sharing.

The Solution: Continuous Optimization

ImproveFast provides the feedback loop your agents need. It runs live experiments, compares variants in real-time, and learns which approaches work best in production — automatically adapting as your agent runs.

Before vs After ImproveFast

Before ImproveFast
×Manual A/B testing takes weeks
×No idea which prompt actually works
×Deploy, test, redeploy cycle
With ImproveFast
✓Learns best variant in hours, not weeks
✓Real production data guides decisions
✓Deploy once, optimize continuously

No Installation. No Configuration. Just Works.

The ImproveFast server is already running. One command to start optimizing.

claude mcp add --transport http improve-fast https://improve.fast/mcp

For Claude Desktop, Cursor, or other MCP clients, see the getting started guide.

▸ 30-second setup
◉ Hosted service
◆ Free to use

When ImproveFast Fits Best

✓
You have multiple variants or strategies to compare.
✓
You get feedback quickly (seconds or minutes).
✓
You want the system to learn as it runs.
!
Avoid if outcomes are slow or safety-critical.

Example Workflow

Agent
→
select
→
ImproveFast
→
run variant
→
record
→
ImproveFast
→
status

Frequently Asked Questions

How long until I see results?

Most experiments converge in 50-150 evaluations. If your agent handles 10 requests/hour, you'll have a winner in 5-15 hours.

Is my data private?

Yes. Experiments are isolated by ID and expire after 7 days. Your experiment data is stored only for the duration needed to run your optimization, then automatically deleted.

What if I have 10 variants?

Thompson Sampling works with any number of variants. More variants need more evaluations (~50 per variant), but the algorithm efficiently focuses on promising candidates.