Claude Code on a $20 plan, without hitting the limit
Developers on Claude Pro who run out of usage before the day ends.
| Cost | $20 a month (Claude Pro); everything else is free and open source |
|---|
What is in it
| Role | Part | Why |
|---|---|---|
| Coding agent | Claude Code Claude Code is now included in Pro ($17/month annual, $20/month monthly) and Max... |
Default to the mid-size model; keep the largest one for planning. |
| Token saver | Headroom Free and open source (Apache-2.0). Compression runs on your machine. |
Compresses logs, test output and search results before they reach the model. |
| Library docs | Context7 Free for individuals with 1,000 API calls a month. Pro is $10 per seat a month with... |
Current docs for the libraries you use, instead of the agent browsing. |
| Skills | mattpocock/skills | Engineering skills that load only when needed, so CLAUDE.md stays short. |
Set it up
npm install -g @anthropic-ai/claude-codepip install "headroom-ai[all]" && headroom wrap claudenpx skills add mattpocock/skills -a claude-code- Add the Context7 MCP server (one command, see its README), and keep CLAUDE.md under about 200 lines.
Trade-offs
Compression saves the most on noisy output (logs, JSON); on plain code edits the gain is smaller. Check /usage to see what actually uses your allowance.
Comments
- Using this stack? Share what worked and what did not.
Other stacks
Open-source agent with a cheap model, paid per token Plan with a frontier model, code with a fast one A coding agent with guardrails for a team