Skip to content

janwilmake/contextarea

Repository files navigation

Context Area - Fast, Smart Prompting for Devs

Goals

Create the best possible Cloud-Based Context Engineering Experience for Developers that can be self-hosted (on Cloudflare)

  • Easy to Share configurations and results with team
  • Good MCP integration
  • Strong Focus on Ease of use and Accessibility
  • Edit Flow > Iterative Messaging
  • URL Expansion with realtime insights (tokens, mediatype)
  • Prompt Check (for OG, but also for missing context)
  • Full Result is URL, each codeblock is URL
  • Painpoint: Prompting workflows are very private, sharing functionality sucks. Visibility is value. Help devs share and monetise their learnings.
  • Creating the best API for stateful chat completions
  • Creating better ways of LLMs using tools
  • Easy interface for chat completions standard
  • X Login -> user-owned results, user payouts.
  • Easy way to create "agents" that have many MCPs (by default)
  • Integration with OpenRouter to make it super generic

Non-goals

We are not aiming to match the UX non-technical folks need in any way.

  • Chat history
  • Inline previews
  • Styles
  • Integrations
  • Uploads
  • Tools

Contents of this repo

  • app - ContextArea itself
  • stream-server - serves the output files as they are generated
  • redirects - redirect from old names lmpify.com and letmeprompt.com
  • contextarea-v0 - old/previous frontend element to drag&drop context and to see tokens in a textarea.

Why Context Area?

contextarea offers several advantages over traditional AI assistants like Claude

Feature Comparison

Feature contextarea Claude Why it matters
Startup Speed ✓ Instant (<100ms) ✗ Slow contextarea is optimised for speed and TTFT. Results are cached.
Performance ✓ Snappy & reliable ✗ Often slow & buggy Consistent, responsive experience without frustrating delays or crashes
URLs as Context ✓ Built-in ✗ Only with MCP Seamlessly reference external textual files (md, json, etc) without having to first toolcall them.
HTML Rendering ✓ Full capability ✗ Limited (no scripts) Complete HTML renders with scripts and full-screen support. Great for prototyping.
Sharing ✓ One-click ✗ Multiple steps, buggy Instantly share prompts and results with a simple URL
Token Efficiency ✓ Incentivizes edit ✗ Designed as chat with history Encourages prompt editing over replies, resulting in less token use and better results with the same model.
Long Output ✓ Up to limit of model API ✗ Limits at ±8k output tokens, continue button is buggy Long outputs can be useful when generating long files.

URL Context

Only urls that return textual results are supported for now. HTML is disabled by design to incentivize users to improve their context.

Some recommend contexts are:

Result format

Any contextarea URL is available as HTML, JSON, or Markdown. Browsers return HTML by default while developer or apis like curl or fetch default to markdown. You can control the output by appending .json/.md/.html or by specifying an accept header.

It's also possible to get a subset of the markdown through ?key=result|context|prompt

Examples:

npx mdapply CLI

Important

Beta

You can use mdapply to apply a response output to your local filesystem. Just run curl "https://contextarea.com/[id]?key=result" -o apply.md && npx mdapply ./apply.md

Try it yourself (this one will create cli.js) in your cwd:

curl "https://contextarea.com/httpsuithubcomj-m8tfk00?key=result" -o apply.md && npx mdapply ./apply.md

'Prompt it' buttons

Allowing users to easily prompt things about your open source library or template can really reduce friction for developers to adopt it.

  • Point to specific contexts that are useful to use (parts of) your library
  • Show how your project was made

You can link from your README, docs, or website to a prompt button using the following code:

HTML:

<a href="https://www.contextarea.com/YOUR_ID" target="_blank">
  <img src="https://b.contextarea.com/YOUR_TEXT" />
</a>

Markdown:

[![](https://b.contextarea.com/YOUR_TEXT)](https://www.contextarea.com/YOUR_ID)

Example:

[![](https://b.contextarea.com/FAQ)](https://www.contextarea.com/httpsuithubcomj-u4l8lj0)

Result:

Any shared links that were previously generated are free to be reached without ratelimit. However, to do new prompts, please be aware that, although I aim to keep the free plan as big as possible, contextarea is not a free service, and as of now, amount of free, unauthenticated, prompts are capped at 5 per hour, and restricted to cheaper models such as OpenAI ChatGPT 4.1 mini. After this users are prompted to add a balance to keep going.

Pricing

  • All previously generated results are cached forever and free for everyone without ratelimits
  • New users that didn't deposit $ with Stripe get 5 free prompts per hour. This may change in the future.
  • After depositing $ through Stripe, users pay the model price + markup when executing new prompts.
  • The markup is 50% markup on top of model price to account for free usage, creator benefits (coming soon), and make this tool sustainable.

About

Cloud-based Dev-focused Context Engineering IDE, Better than Claude

Resources

License

Contributing

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors