# We Made Claude Pro Last 26.5% Longer

**Sacha Morard**  
co-founder co-CEO  
March 19, 2026  
[Performance](/content/blog/categories/performance/index.html)

You're mid-task. Claude is in flow. Then the plan limit hits and everything stops. You know the feeling — the session cuts out, the context is gone, and you're starting over. For heavy Claude Code users, this isn't an occasional annoyance. It's a regular ceiling on what you can get done in a day.

We built Edgee's Claude Code Compressor to push that ceiling back. And last week, we put it to the test.

## The Endurance Test

We ran a controlled benchmark using our open-source [claude-compression-lab](https://github.com/edgee-ai/claude-compression-lab). The setup was simple:

- Two isolated Claude Code sessions: one running normally, one routed through Edgee's token compressor  
- The same fixed set of 27 coding instructions, pasted one by one into each session  
- Each session tracked plan consumption via `.claude/.claude.json`  
- Both sessions ran until their Claude Pro plan was exhausted

One session stopped at 21 instructions completed. The other kept going.

We extended Claude Pro's limit by 26,5% - YouTube

[We extended Claude Pro's limit by 26,5%](https://www.youtube.com/watch?v=EQHYskM0nmg)

## The Results

| Metric                          | Baseline Claude | Claude + Edgee | Improvement                      |
| ------------------------------- | ---------------- | --------------- | ---------------------------------| 
| Instructions completed          | 21               | 26.5           | **+26.2%**                       |  
| Plan consumed per instruction    | 4.76%           | 3.77%          | **20.8% more efficient**         |  
| Total session cost              | $10.25          | $12.26         | +19.6%                           |  
| Cost per instruction            | $0.488          | $0.463         | **5.1% cheaper per task**       |

The Edgee session spent more in absolute terms — because it completed more work. But on a per-task basis, it was actually 5% cheaper. The plan stretched further and went further.

_Same plan. Same tasks. 26.5% more work done._

## How It Works

Edgee sits between Claude Code and Anthropic's API. Before each request is sent, Edgee compresses the conversation history and context — stripping redundancy while preserving meaning. Claude receives a smaller, cleaner prompt. Less token overhead means each request consumes less of your plan, and the same fixed budget covers more ground.

The key challenge in compression is preserving semantic fidelity: remove too much and you degrade the response. Edgee handles this automatically, applying compression policies tuned for Claude Code's patterns. For a deeper look at the technique, read [Achieving More With Less Using Token Compression](/content/blog/posts/2026-02-19-achieving-more-with-less-using-token-compression/index.html).

## What This Means For You

If you're a heavy Claude Code user hitting plan limits regularly, this is a direct quality-of-life improvement. No workflow changes required — you just point Claude Code at Edgee's gateway instead of Anthropic's API directly as explained in this [documentation](/content/docs/features/claude-compression/index.html).

The math is straightforward: if your current plan runs out before your work does, you either pay for more plans or you stop. Edgee lets you do more with the plan you already have.

## What's Next

These are our first published results, and there's more headroom ahead. We're actively improving our compression models to push efficiency further — the 26.5% gain is a floor, not a ceiling. We're also building compressors for the other popular coding assistants, so Claude Code won't be the only one to benefit for long.

* * *