MCP Tools DirectoryAbout this integration
Statistical regression testing for LLM agents: p-value, effect size, and CI on behavior change.
- Transport
- stdio
- Authentication
- none
- Initial setup
- none
- Runtime
- unattended
- Evidence
- documented
- Version
- 0.1.8
- Package
- agent-regress-cli
- Last compatibility test
- Not independently tested
Connect your agent
agent-regress-cliPublisher documentation and the executable manifest were reviewed; the package was not executed or independently security-audited. The MCP tool shells out to the local agent-regress CLI and can read caller-selected result files, so filesystem and client approval boundaries still apply. Framework extras and user-supplied agents can have their own credentials, but the listed MCP artifact does not manage them.
Capabilities: statistical agent regression testing, p-value and effect-size reporting, bootstrap confidence intervals, JSON CLI execution
Connected profiles
Additional details
io.github.RudrenduPaul/agent-eval
Source ↗ · Checked 2026-09-170.1.8
Source ↗ · Checked 2026-09-17CC0-1.0; package licenses are separate
Source ↗ · Checked 2026-09-17pypi
Source ↗ · Checked 2026-09-17agent-regress-cli
Source ↗ · Checked 2026-09-190.1.9
Source ↗ · Checked 2026-09-19The publisher documents that an MCP client spawns the local server subprocess and can run comparisons without a human invoking the CLI for each request.
Source ↗ · Checked 2026-09-19The MCP server operates on caller-supplied local evaluation inputs; the publisher states that the package does not make model calls and has no credentials to configure for its statistical core.
Source ↗ · Checked 2026-09-19The agent-regress-mcp console script is launched by an MCP client as a local subprocess using stdio transport.
Source ↗ · Checked 2026-09-19Publisher documentation reviewed; package and integration endpoint not executed or independently security-audited.
Source ↗ · Checked 2026-09-19