HTMLBuild AI agents that keep working.
Test MCP servers, manage the local stack, lock agent dependencies, and turn repeated work into reusable systems.
- 01ConfigureDefine stack, tools, and permissions.
- 02VerifyRun compatibility scans.
- 03LockPin agent dependencies.
- 04ImproveMine repeated work.
{
"mcp": {
"servers": [
{ "name": "workspace", "transport": "stdio" },
{ "name": "flightcheck", "transport": "https" }
],
"permissions": { "network": "limited" },
"timeout_ms": 30000
}
}One lifecycle.
Fewer failures.
Move from idea to reliable agent with a repeatable system. Each stage produces evidence and context your team can trust.
AI Deck
Inventory local AI CLIs and MCP client configurations from one private Windows control center.
MCP FlightCheck
Test the handshake, discovery surface, schema budget, and client compatibility signals of a public MCP server.
AgentLock
Commit a deterministic record of the dependencies that influence agent behavior and detect drift in CI.
Workflow Miner
Find repeated prompts and tool chains, then turn the evidence into skills, connectors, or automations.
Start with the free AI Tools Lab.
Measure MCP schema overhead, convert client configurations, find tool collisions, and budget a model context locally in your browser.
111 private browser tools remain free.
Format, convert, inspect, encode, generate, and debug without uploading classic transforms to a server.
Use the same deterministic tools from browser, API, or MCP.
The free toolbox handles one-off work. Paid plans add API batches, the hosted MCP connector, extension workflows, higher limits, shared operations, and the new AI developer products.
From the workbench
Practical engineering notes for web and AI builders.