Reduce AI agent token waste
by 60–90%
Every MCP server adds tool definitions to your AI agent's context. ContextOpt filters them so only relevant tools consume tokens.
$ npx contextopt analyzeWorks with
Claude Code
Cursor
Windsurf
Antigravity
OpenCodeClaude CodeCursorWindsurfAntigravityOpenCodeEvery MCP server adds 10–30 tools to your context
100+
tool definitions
5 MCP servers can add over 100 tool schemas to every prompt your AI agent sends.
20K+
extra tokens
That's 20,000+ tokens of tool metadata loaded into context on every single message.
60–90%
reduction possible
ContextOpt filters tools by profile. Only the ones you actually need consume context.
How it works
1
Analyze your config
$ npx contextopt analyze2
Pick a profile
$ npx contextopt profile -s coding3
Start the proxy
$ npx contextopt start --profile coding4
Connect your AI agent
$ npx contextopt start --http --port 3001What you save with ContextOpt
Pick the scenario closest to you. Your token savings pay for Pro many times over.
Solo Dev
25 sessions/day · 3 servers
Power User
50 sessions/day · 5 servers
Small Team
200 sessions/day · 8 servers
The typical developer saves $72/month on tokens, pays \$20/month for Pro, and keeps $53/month in their pocket.
Calculate your savings
Token cost / mo
$56
You save
$39
Net profit
+$20/mo
After $39 savings minus $19 Pro
Upgrade to Pro — Save $39/moFree vs Pro
Free
$0
- ✓CLI analyze & profile
- ✓Stdio proxy
- ✓Basic token estimation
- ✗HTTP proxy
- ✗Advanced analytics
- ✗Priority support
Pro
\$20/month
- ✓Everything in Free
- ✓HTTP proxy (--http)
- ✓Advanced token analytics
- ✓Profile comparison reports
- ✓Priority email support
Ready to optimize your context?
$ npx contextopt analyze