io.github.physics-star-cat/databutler-stats

Measured conclusion: Deep-tested. L2 outcome: tools_ok.

Exact statistics & probability: distributions, hypothesis tests, CIs, Bayesian updates, regression.

Add to your agent

Copy this into your MCP client config:

{
  "mcpServers": {
    "databutler-stats": {
      "url": "https://databutler.dev/api/mcp/stats",
      "transport": "streamable-http"
    }
  }
}

Measured tools (6)

• descriptive_stats read-only

  Summary statistics for a numeric array: mean, median, sd, variance, quartiles, IQR, skewness, min/max.

  Parameters:

  • data (required) — array

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• distribution read-only

  Evaluate a probability distribution (normal, t, chi2, binomial, poisson): pdf/pmf and cdf at a value, and/or the quantile at a probability, plus mean & variance. Params per dist: normal {mean,sd}, t {df}, chi2 {df}, binomial {n,p}, poisson {lambda}.

  Parameters:

  • dist (required) — string
  • p — number — probability to get the quantile for (0-1)
  • at — number — value to evaluate pdf/pmf and cdf at
  • params — object

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• hypothesis_test read-only

  Run a significance test and get the statistic, p-value, and a plain-language interpretation with assumptions. test = one-sample-t {data, mu0}, two-sample-t {data1, data2}, one-proportion-z {successes, n, p0}, two-proportion-z {successes1,n1,successes2,n2}, chi2-gof {observed, expected?}, chi2-independence {table}. Optional tail: two-sided (default) | greater | less; alpha default 0.05.

  Parameters:

  • test (required) — string
  • n — number
  • n1 — number
  • n2 — number
  • p0 — number
  • mu0 — number
  • data — array
  • tail — string
  • alpha — number
  • data1 — array
  • data2 — array
  • table — array
  • expected — array
  • observed — array
  • successes — number
  • successes1 — number
  • successes2 — number

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• confidence_interval read-only

  Confidence interval for a mean (t-based; from data, or n/mean/sd) or a proportion (Wilson; successes/n). kind = mean | proportion; confidence default 0.95.

  Parameters:

  • n — number
  • sd — number
  • data — array
  • kind — string
  • mean — number
  • successes — number
  • confidence — number

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• bayes_update read-only

  Discrete Bayesian update: given competing hypotheses each with a prior and the likelihood of the observed evidence, return normalised posteriors. Priors are renormalised to sum to 1.

  Parameters:

  • hypotheses (required) — array

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• linear_regression read-only

  Simple linear regression of y on x: slope, intercept, r, r², slope std error and p-value, equation.

  Parameters:

  • x (required) — array
  • y (required) — array

  Hints: read-only yes · destructive no · idempotent yes · open-world no

Publisher website Repository

Measured by MCPYP probing. Untested fields are omitted, never estimated. Listing is always free; MCPYP never proxies calls or stores user tokens.