Playwright-based testing and eval framework for MCP servers with LLM-as-a-judge
This repository also configures its own agents. See what mcp-server-tester tells them →
Skill Claude CodeCodex
Reference for @gleanwork/mcp-server-tester (MST) 2.0, the Playwright-based testing and evaluation framework for MCP servers, with the mst CLI. Covers the vocabulary (test, eval, case, client, variant, trial, assertion, judge), entry points, server config and auth, the mcp fixture and matchers, eval datasets and eval…
Skill Claude CodeCodex
Improve an MCP server's tool metadata (tool names, descriptions, input schema descriptions) with measured evidence, using runToolOptimization from @gleanwork/mcp-server-tester/evals or mst run with tool-metadata variants. Use when asked to improve tool descriptions, fix tools a model misses or over-uses, optimize tool…
Skill Claude CodeCodex
Write evals for MCP servers with @gleanwork/mcp-server-tester (MST). An eval gives a client and model the cases of a dataset (an input, what is expected, and assertions) and grades each trial's trace. Run the cases in a Playwright test with runEvalDataset on the mst client, or with mst run and an eval config that…
Skill Claude CodeCodex needs its repo
Write Playwright tests for MCP servers with @gleanwork/mcp-server-tester (MST). A test calls a server's tools (or sends any MCP request) directly through MST's fixtures and checks the result with its matchers, with no client or model involved. Use when asked to write MCP tests, test MCP tools, check tool responses…
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: