Works with the AI Agents you already use
token-budget — AI Context Budget Optimizer
Diagnose supplied context with token-budget MCP, compare eviction strategies, trim tool text, and preserve coding-session checkpoints. Estimates only; no automatic host-context control.
$1
token-budget — AI Context Budget Optimizer
Example session with this skill installed
Analyze a synthetic sample under maxTokens 400 and reserve 100: pinned system 'Preserve requirements.', user 'Diagnose context pressure.', and tool text 'build log ' repeated 200 times.
- Read your context and instructions
- Compiled the token-budget
Source-built stdio MCP test returned 525 estimated message tokens for 3 messages against an effective scenario budget of 300. Tool text plus framing contributed 504 estimated tokens (96% of supplied context); pinned requirements contributed 10. Warnings: HIGH_UTILIZATION, TOOL_OUTPUT_DOMINANT, OVER_BUDGET. A separate truncate_tool_output experiment with keep=both reduced the 500-estimated-token text to 80. This synthetic example demonstrates behavior, not real-world savings. Narrow future output at the source; truncation cannot reclaim content already read by the host.
Connects securely to your tools. The creator never sees your data.
About this skill
Original MCP operating workflows for token-budget: diagnostics on supplied message samples, four-policy comparisons, selective tool-output handling and coding-session checkpoints. Includes exact schemas, setup guides, client examples and failure handling.
Requires the separate open-source token-budget MCP server and Node.js >=18. The server remains free under MIT. This package is MIT licensed; no exclusive rights are promised. Source-built execution verified: all 14 tools exercised and 298 core/MCP tests passed. IDE setup is documentation-verified, not individually executed.
Release limitations: source lockfile audit reports 8 production dependency advisories (1 critical, 4 high, 3 moderate). Remediation and retesting are required before production use. Registry launch timed out here; source build is the verified route. MCP is a testing/diagnostic surface, not automatic host context management. All token measurements are estimates for supplied data, not provider bills or hidden live context.
No automatic host compaction, hidden/live context meter, provider billing accuracy, summarizer callback or session persistence. get_context does not replace raw history. Runtime audit has unresolved advisories; not production-cleared. MIT package rights need reconciliation with marketplace buyer restrictions before publication.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 19 days ago
Requires
Permissions
Allowed Hosts
Network/terminal for optional runtime installation; local diagnostics require no network or provider API key. Read only task-authorized files. Write only authorized checkpoints. Remote content transmission requires destination-specific authorization.