# Toolport Boosts AI Agent Performance Local gateway solution reduces token usage by up to 90% - **Source:** [Product Hunt — The best new products, every day](https://brifing.si/source/a459d489-2e9f-45f3-a76f-4f40ae5ec434) - **Published:** 2026-08-07 - **Topic:** [Data & AI](https://brifing.si/c/data-ai) - **Original article:** - **On Brifing:** https://brifing.si/s/3ff26ba2-7273-42ab-9865-64fa70f0fa1a ## Problem: Excessive Token Usage MCP servers send entire tool lists with each request, causing massive token waste. For example, 3 servers consume 24,000 tokens just for tool definitions per request. This significantly degrades AI agent performance. Developer Tyler created Toolport to solve this exact problem. ## Solution: How Toolport Works Toolport acts as a local gateway that exposes meta-tools instead of sending all definitions at once. Agents search for tools only when needed. This approach reduces definition overhead by 99.5% in a 415-tool catalog. It maintains the same task success while cutting token usage by up to 90%. ## Security Features Toolport offers both performance and security. It fingerprints each tool to detect rug-pulls and tool poisoning attacks. Destructive calls require user approval - like table deletions. API keys stay in the OS keychain, never in plain text client configs. Injection attempts get flagged automatically. ## Scalability Advantage In a 14-server/415-tool setup, traditional methods consume ~165,000 tokens per request while Toolport reduces this to 886 tokens. Even when using 9 different tools, total cost remains ~3,600 tokens. The advantage becomes more pronounced as systems scale. ## Platform Support & License Toolport runs on Windows, macOS, and Linux. It's MIT-licensed and fully open source. A headless Docker version is available. Currently supporting 33 client apps, with new additions based on user feedback. The roadmap prioritizes community requests. ## Advanced Features Updated in July 2026, Toolport supports the latest MCP spec. Code mode lets agents run tool sequences as server-side scripts. Approval binds to exact arguments seen during review. The Teams version ensures keys never leave the machine. All traffic stays local with no routing through external servers. ## Developer Experiences Developers like Abdullah Javaid report similar challenges in Claude Code. Toolport's cost model scales with usage rather than catalog size. Tyler's benchmarks provide concrete data showing the system's advantages. The community actively discusses optimization tradeoffs. ## Impact on AI Ecosystem Toolport enables AI agents to handle complex tasks more efficiently. Being open source and local accelerates adoption. Its security features make it enterprise-ready. As token efficiency becomes crucial, Toolport emerges as a vital productivity tool for AI developers working with multiple MCP servers. ## Sources covering this story - https://www.producthunt.com/products/conduit-12 — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._