HyphAeon from Claude

Model Context Protocol server · @veg/hyphaeon-mcp · stdio and remote

The same analyses as a Model Context Protocol server. Locally it runs on your machine with the vendored model, reads files by path, and sends nothing anywhere; remotely it is a claude.ai connector on this site's origin. One tool, hyphaeon_analyze, does what dropping a file on the front page does and returns the report; the per-pillar tools are there for a single question.

Local, over stdio

For Claude Code, one line:

claude mcp add hyphaeon -- npx @veg/hyphaeon-mcp

Or start the server yourself and point any MCP client at its stdio:

npx @veg/hyphaeon-mcp

Current release: @veg/hyphaeon-mcp 0.5.0 — read from the package at build, with the 15 tools listed below.

The server runs the ONNX graphs under Node with onnxruntime-node; the model files ship with the package and the hash of each graph is verified before it scores. Over stdio, alignment, tree and the other file inputs accept a file:// URL, so a transcript never carries the sequences. Results write provenance.surface: "mcp-stdio".

HYPHAEON_MCP_THREADS: onnxruntime intra-op threads (default 1; 4 to 8 on a laptop).
HYPHAEON_MODELS_DIR: a directory with manifest.json and the graphs, to use a different build of the model.

Nothing else is needed on the host: no Python, no HyPhy. Every pillar, phenotype included, runs in this process, and a tree is optional — without one, or without branch lengths, the run takes pairwise TN93 distances (the reference's --use-tn93).

Remote, as a claude.ai connector

Add a custom connector with this site's /mcp endpoint. Authorization is OAuth 2.1 with dynamic client registration, PKCE, and an out-of-band redirect for headless clients, auto-approved the way the Datamonkey connector is, so the ceremony completes without an account.

https://<host>/mcp

The remote server is the same code under the same caps as the Node server: your sequences are sent to it, it says so in the tool description, keeps nothing in logs, expires each job after 7 days, and refuses file:// inputs. Results write provenance.surface: "mcp-http". The endpoint ships with the Node server; the host is the deployment decision the plan records as D1.

Tools

Every analysis tool runs in-process: the JavaScript port runs inside the MCP process and the result matches the reference at the published parity classes. That has been true of all eight since Phase 3 — hyphaeon_phenotype included, which was the last tool to answer through a Python subprocess — so nothing here is marked "bridged" and no tool result carries provenance.surface: "python-reference". The names below are the ones @veg/hyphaeon-mcp 0.5.0 registers.

**Tools.** One row per tool the server registers, `hyphaeon_analyze` first; what each returns.
Tool	Returns
`hyphaeon_analyze`	The whole report: diagnostics, sites, gene, epistasis and sectors, attribution, filter, DMS (progressive, capped), phenotype when a trait is given, provenance and timings, in the schema the web report reads. The same options as the report's "Re-run with": model_variant, max_species, reference_sequence, call_mode, seed, permutations, dms; plus phenotype / phenotype_file for the trait, and use_tn93 (a tree is optional).
`hyphaeon_validate`	Diagnostics with the same warning codes as the browser, the run mode (inline or job) and a cost estimate.
`hyphaeon_meme`	Per-site LRT, p, q, invariable flag and the report's rank columns; --filter and --attribute as options.
`hyphaeon_busted`	p_ACAT, p_Simes, omnibus LRT, selection energy, significant-site counts, and the neural head's fields.
`hyphaeon_epistasis`	Co-selection edges with CESI and q, sectors with coherence and p_perm, optional per-sector DMS, GraphML.
`hyphaeon_dms`	19-substitution scan per site with intrinsic plasticity and the ΔLRT map.
`hyphaeon_phenotype`	Trait association per site, PARS signature, trait sectors, permulation p. The trait is a preset, a foreground list or pattern, or an inline trait table (phenotype_file); permulations need a tree with branch lengths.
`hyphaeon_dates`	The date review stage as data: which rule dated each sequence, what did not match, what was imputed, and whether the set carries a clock at all. Loads no model and no graph — milliseconds. Call it first: it reports the two gates the two pillars below refuse on (dates read mostly as bare numbers in a sequence name, and sequences that carry none).
`hyphaeon_dating`	The molecular clock: rate, MRCA date with its interval, the spline adjudication, and a per-taxon table of residuals and outliers. Model-free by default, and takes no tree (D34). use_model adds the second ONNX artifact and a different estimator — on the bundled korber example the two ancestor dates are twelve years apart, so the mode is quoted with the date.
`hyphaeon_temporal`	Per-site selection trajectories through calendar time, a permutation null, four fPCA wave modes, and a four-way classification against the static call. Always a job; the record is megabytes and is read one section at a time. Every section carries the honesty block: which of the null's four states this run is in, that p_perm is 1.0 at untested codons by the reference's own fill, and that the wave shares move with the null.
`hyphaeon_evaluate`	Concordance of a meme CSV with a HyPhy MEME JSON: correlations, ROC-AUC, PPV, FPR, confusion matrices.
Job control
`job_status`	Phase, progress and warnings of a queued run.
`get_results`	A completed job's result, shaped with fields, top or summary_only.
`cancel_job`	Cancels a queued or running job.
`list_models`	The weights manifest and the engine's status.

Every analysis tool accepts fields, top and summary_only to shape its result, and answers inside the call when the alignment is at most 12,000 codons and the work term sites × taxa² is at most 2.5 × 10⁹ (× 19 for DMS); above that, or with run_async, it returns a job id for job_status and get_results. Errors are {error, kind, hint} with kind either input (your alignment, tree or options) or server (the model files or the runtime).

Resources: hyphaeon://models, hyphaeon://methods/requirements, hyphaeon://caveats (the same numbers as the methods page), hyphaeon://examples/{name}, hyphaeon://gallery. Prompts: one interpretation guide per pillar, which is how the closing turn of the transcript below knows what to say.

Example transcript

Recorded 2026-09-12 against @veg/hyphaeon-mcp 0.5.0 over stdio, on Smc6 (20 primates × 1,097 codons) and H5N1_HA_geo (98 sequences × 566 codons, with its tree and metadata CSV), both bundled examples. Tool results are abridged to what a reader needs; the numbers are the server's.

You
Is there episodic selection in Smc6? The alignment is examples/Smc6.fasta and the tree examples/Smc6.nwk.

hyphaeon_validate · 98 ms

{
  "alignment": "file:///path/to/examples/Smc6.fasta",
  "tree": "file:///path/to/examples/Smc6.nwk"
}

Result

{
  "ok": true,
  "warnings": [
    { "code": "NON_ACGT_FRACTION", "severity": "info",
      "message": "0.4% of characters are not A/C/G/T (249 gaps, 2 ambiguity or other codes)." },
    { "code": "SHALLOW_TREE", "severity": "info",
      "message": "Median patristic distance 0.0238 substitutions/site is shallow (< 0.05); the viral variant was trained on this regime.",
      "data": { "suggestVariant": "viral" } },
    { "code": "COST_ESTIMATE", "severity": "info",
      "message": "1097 codons x 20 taxa: about < 1 s of model time on a laptop CPU (reference path)." },
    { "code": "RUN_MODE", "severity": "info",
      "message": "hyphaeon_meme runs in-process (ONNX Runtime under Node) and answers inside the tool call." }
  ],
  "summary": { "format": "fasta", "sequence_count": 20, "taxa_matched": 20, "unique_haplotypes": 20,
               "codons": 1097, "tree_source": "user", "match_tier": "exact", "tree_free": null,
               "branch_lengths": { "branches": 37, "missing": 0, "zeros": 2 }, "mode": "sync" }
}

hyphaeon_meme · 243 ms

{
  "alignment": "file:///path/to/examples/Smc6.fasta",
  "tree": "file:///path/to/examples/Smc6.nwk",
  "summary_only": true
}

Result

{
  "analysis": "meme",
  "summary": {
    "taxa_count": 20, "codon_count": 1097, "runtime_sec": 0.153,
    "invariable_sites": 1000, "variable_sites": 97,
    "significant_p05": 1, "significant_p10": 5, "significant_q05": 0, "significant_q10": 0,
    "top_sites": [
      { "site": 697, "hyphaeon_lrt": 6.913, "p_value": 0.0141, "q_value": 0.667, "call": "Top 2%", "percentile": 100 },
      { "site": 930, "hyphaeon_lrt": 4.068, "p_value": 0.0611, "q_value": 0.667, "call": "Top 2%", "percentile": 99.0 },
      { "site": 628, "hyphaeon_lrt": 3.735, "p_value": 0.0726, "q_value": 0.667, "call": "Top 5%", "percentile": 97.9 },
      { "site": 365, "hyphaeon_lrt": 3.405, "p_value": 0.0863, "q_value": 0.667, "call": "Top 5%", "percentile": 96.9 },
      { "site": 279, "hyphaeon_lrt": 3.292, "p_value": 0.0916, "q_value": 0.667, "call": "Top 5%", "percentile": 95.9 }
    ]
  },
  "collections": { "sites": 1097 },
  "provenance": {
    "surface": "mcp-stdio", "engine": "in-process", "model_version": "v1", "model_variant": "general",
    "artifact_sha256": "aa10e8e0f059436bd8a6f59e1fb3f0137e845533b9c65009230c7de6b91ef22c", "artifact_verified": true,
    "is_surrogate": true, "surrogate_for": "MEME", "seed": 42, "elapsed_sec": 0.235,
    "preprocessing": { "taxa_used": 20, "duplicates_collapsed": 0, "pd_subsampled": false,
                       "tree_source": "user", "tree_provided": "user", "tree_free": null,
                       "display_tree_source": "user", "distance_rescaled": false },
    "reference_command": ["hyphaeon", "meme", "-a", "Smc6.fasta", "-t", "Smc6.nwk", "--cpu", "-o", "<out.json>"]
  }
}

hyphaeon_busted · 117 ms

{
  "alignment": "file:///path/to/examples/Smc6.fasta",
  "tree": "file:///path/to/examples/Smc6.nwk",
  "summary_only": true
}

Result

{
  "analysis": "busted",
  "summary": {
    "gene": "Smc6", "taxa": 20, "sites": 1097,
    "p_value_acat": 0.1180, "p_value_simes": 1.0, "omnibus_lrt": 3.299, "total_selection_energy": 84.82,
    "sig_sites_p05": 5, "sig_sites_p10": 14,
    "selection_probability": 0.533, "predicted_gene_lrt": 0.330, "synonymous_rate_variation": 0.607,
    "rate_distributions": { "omega_1": 0.1, "proportion_1": 0.101, "omega_2": 1, "proportion_2": 0.738,
                            "omega_3": 1.164, "proportion_3": 0.161 },
    "positive_selection_detected": true
  },
  "provenance": {
    "surface": "mcp-stdio", "surrogate_for": "BUSTED", "seed": 42, "elapsed_sec": 0.113,
    "busted_head_sha256": "2ad554e02b5580b813a915d8fbd8a4fa925c7679977adb596388fdc8d4c1c116", "busted_head_verified": true,
    "neural_head": { "deterministic_upstream": false }
  }
}

You
Do the top sites track the great apes specifically, rather than the primates in general?

hyphaeon_phenotype · 832 ms

{
  "alignment": "file:///path/to/examples/Smc6.fasta",
  "tree": "file:///path/to/examples/Smc6.nwk",
  "foreground": "hg18,homSap_293T,panTro4,panPan,ponAbe2,nomLeu3",
  "permulations": 100,
  "seed": 42,
  "summary_only": true
}

Result

{
  "analysis": "phenotype",
  "summary": {
    "phenotype_meta": { "mode": "discrete", "foreground_count": 6, "background_count": 14,
      "description": "User-specified foreground patterns: ['hg18', 'homSap_293T', 'panTro4', 'panPan', 'ponAbe2', 'nomLeu3']" },
    "taxa_count": 20, "codon_count": 1097, "sites": 97,
    "spectral_energy": 0.5557, "norm_spectral_ratio": 0.3251,
    "max_assoc": 0.99999824, "p_evd_length_adjusted": 0.00846,
    "compact_pars_signature": "[ R704H - Q567P - D930E - K763E - A697A - N913S - M685T - ... ]",
    "permulations_count": 100, "gene_p_value_perm": 0.485,
    "significant_sites_count": 0, "trait_sectors": 0, "coselection_pairs": 0,
    "top_sites": [
      { "site": 704, "ref_aa": "R", "derived_aa": "H", "hyphaeon_lrt": 1.861, "association_rho": 1.000, "score": 1.364, "foreground_freq_pct": 100, "background_freq_pct": 0 },
      { "site": 567, "ref_aa": "Q", "derived_aa": "P", "hyphaeon_lrt": 1.536, "association_rho": 1.000, "score": 1.240, "foreground_freq_pct": 100, "background_freq_pct": 0 },
      { "site": 930, "ref_aa": "D", "derived_aa": "E", "hyphaeon_lrt": 4.068, "association_rho": 0.596, "score": 1.202, "foreground_freq_pct": 100, "background_freq_pct": 0 },
      { "site": 763, "ref_aa": "K", "derived_aa": "E", "hyphaeon_lrt": 2.091, "association_rho": 0.816, "score": 1.181, "foreground_freq_pct": 66.7, "background_freq_pct": 0 },
      { "site": 697, "ref_aa": "A", "derived_aa": "A", "hyphaeon_lrt": 6.913, "association_rho": 0.434, "score": 1.141, "foreground_freq_pct": 66.7, "background_freq_pct": 35.7 }
    ]
  },
  "provenance": {
    "surface": "mcp-stdio", "engine": "in-process",
    "is_surrogate": true, "surrogate_for": "no HyPhy counterpart (attention-based trait association)",
    "seed": 42, "elapsed_sec": 0.830,
    "options": { "foreground": "hg18,homSap_293T,panTro4,panPan,ponAbe2,nomLeu3", "permulations": 100, "seed": 42 },
    "preprocessing": { "taxa_used": 20, "tree_source": "user", "tree_free": null },
    "reference_command": ["hyphaeon", "phenotype", "-a", "Smc6.fasta", "-t", "Smc6.nwk", "--cpu",
                          "--foreground", "hg18,homSap_293T,panTro4,panPan,ponAbe2,nomLeu3",
                          "--permulations", "100", "--seed", "42",
                          "--mds-sign", "canonical", "-o", "<out.json>"]
  }
}

Claude
Smc6 has 1,097 codons, 97 of them variable across the 20 primates. One site reaches nominal p ≤ 0.05, codon 697 (LRT 6.9, p 0.014), with codons 930, 628, 365 and 279 next in rank; no site survives false-discovery control (every q is 0.67), so at the gene level the combination tests are neutral (p_ACAT 0.118, p_Simes 1.0, omnibus LRT 3.3). The "positive selection detected" flag is true only because the neural head's selection probability is 0.53, and that field is not reproducible upstream, so I would not lean on it. On the great-ape question: several sites do separate the six great apes cleanly (704 R→H and 567 Q→P are present in all six and absent from the other fourteen, ρ ≈ 1.0), but with a Brownian-motion permulation null over the tree the gene-level p is 0.49 — that pattern is what phylogenetic autocorrelation alone produces for a clade-shaped trait on this tree, and no site survives with a permulation p. Two things to keep in mind throughout: these are surrogate scores for MEME (rank is reliable, the LRT scale is compressed), and the tree is shallow (median patristic distance 0.024), the regime the viral variant was trained on, so a re-run with model_variant "viral" is worth comparing. If codon 697 matters to you, confirm it with MEME on Datamonkey.
You
Different dataset: I have 98 avian influenza HA sequences sampled over about a decade (examples/H5N1_HA_geo.fasta, tree H5N1_HA.nwk, dates in H5N1_HA_metadata.csv). Did anything sweep, and when?

hyphaeon_dates · 25 ms

{
  "alignment": "file:///path/to/examples/H5N1_HA_geo.fasta",
  "dates_file": "file:///path/to/examples/H5N1_HA_metadata.csv",
  "rows": false
}

Result

{
  "analysis": "dates",
  "ok": true,
  "headline": "98 of 98 sequence(s) dated from table — 98 by numeric; 1996 to 2005 (years). Units years (inferred).",
  "clock": { "has_clock": true, "dating_possible": true, "temporal_possible": true, "reasons": [], "temporal_reasons": [] },
  "gate": { "ok": true, "blocking": [], "overrides": { "accept_bare_numbers": false, "drop_undated": false }, "applied": [] },
  "date_review": {
    "source": "table", "time_units": "years", "time_units_source": "inferred",
    "coverage": { "taxa_total": 98, "dated": 98, "undated": 0, "coverage": 1, "from_table": 98, "from_header": 0, "imputed": 0 },
    "by_rule": { "numeric": 98 },
    "span": { "min": 1996, "max": 2005, "span": 9, "unique": 10, "tied": 29 },
    "match_tiers": { "exact": 98, "case_insensitive": 0, "first_token": 0, "field_containment": 0 },
    "table": { "delimiter": ",", "delimiter_source": "sniffed", "columns": ["taxon", "date", "location", "lat", "lon"],
               "strain_col": "taxon", "date_col": "date", "rows_read": 98, "rows_dated": 98, "duplicates": [] },
    "unmatched_metadata": { "count": 0, "names": [] },
    "warnings": [
      { "code": "DATES_DELIMITER_GUESSED", "severity": "info" },
      { "code": "DATES_UNITS_INFERRED", "severity": "warn" }
    ]
  },
  "engine": "in-process (no model, no graph)"
}

hyphaeon_dating · 85 ms

{
  "alignment": "file:///path/to/examples/korber_env_gp160.fasta",
  "root_taxon": "CONSENSUS",
  "drop_undated": true,
  "summary_only": true
}

Result

{
  "analysis": "dating",
  "summary": {
    "taxa_count": 142, "dated": 142, "timespan": [1959.5, 1997.5],
    "distance_mode": "tn93", "active_model": "spline", "clock_model": "auto", "ci_method": "fieller",
    "selected_clock": "Restricted Spline (rate deceleration (0.04x) detected: F=5.26, p=0.0234, dAIC=+3.3)",
    "t_mrca": 1938.77, "ci_mrca": [1938.77, 1938.77], "mu": 2.2627e-3,
    "root_description": "explicit_root_CONSENSUS", "latent_root": null,
    "ols": { "mu": 1.1690e-3, "t_mrca": 1893.91, "ci_fieller": [1850.90, 1916.79], "r2": 0.2314, "n": 141 },
    "pgls": null,
    "outliers": 0, "taxa_summary_rows": 142,
    "honesty": {
      "distance_mode": "tn93", "model_pass": false,
      "estimators_not_built": [ "Attention PGLS", "Latent root search", "Power-law clock", "Leave-one-out / jackknife" ],
      "note": "t_mrca moves with distance_mode: model-free and model-based are different answers, measured twelve years apart on this example (1938.77 against 1926.81)."
    }
  },
  "provenance": {
    "surface": "mcp-stdio", "engine": "in-process", "elapsed_sec": 0.083,
    "preprocessing": { "tree_source": "tn93", "date_source": "header", "date_units": "years",
                       "date_gate": { "ok": true, "applied": ["DATES_UNDATED_PRESENT"] } },
    "reference_command": {
      "command": "hyphaeon dating -a alignment.fasta --no-tree --distance-mode tn93 --root-taxon CONSENSUS --method ols --clock-model auto --ci-method fieller --cpu -o <out.json> -c <out.csv>",
      "reproduces": false,
      "caveats": [
        "Running this does not make the two files diff clean ... compare numerically, not with diff.",
        "The dates on this run were read from the sequence names by this build's own date layer, which reads headers hyphaeon dating cannot (142 of 143 by the korber_isolate rule). A different dated set is a different regression; export the dates as a two-column CSV and pass -d to reproduce this run.",
        "This build does not estimate Attention PGLS, Latent root search, Power-law clock, Leave-one-out / jackknife, so no flag for them is printed above."
      ]
    }
  }
}

hyphaeon_temporal · 4291 ms

{
  "alignment": "file:///path/to/examples/H5N1_HA_geo.fasta",
  "tree": "file:///path/to/examples/H5N1_HA.nwk",
  "dates_file": "file:///path/to/examples/H5N1_HA_metadata.csv",
  "dates_file_name": "H5N1_HA_metadata.csv",
  "n_permutations": 1000,
  "time_points": 250
}

Result

{
  "analysis": "temporal", "job_id": "4368ab5e04a952f42fce650bcd9ff05c", "status": "completed", "stage": "complete",
  "sections": ["summary","sites","curves","waves","permutations","dates","candidates","warnings","honesty","provenance"],
  "summary": {
    "taxa_total": 97, "taxa_timestamped": 97, "codons_total": 566, "codons_variable": 169,
    "timespan_years": 9, "t_min": 1996, "t_max": 2005, "bandwidth_years": 0.45,
    "sig_static_q10": 1, "stage1_candidates": 168,
    "confirmed_sweeps": 16, "concordant_sweeps": 0, "rescued_sweeps": 16, "filtered_static_noise": 1,
    "fpca_wave_variance_pct": [97.13, 1.48, 0.60, 0.47], "runtime_sec": 4.018,
    "regime": { "time_units": "years", "sweep_mode": "episodic", "prune_duplicates": true },
    "dates": { "file": "H5N1_HA_metadata.csv", "dated": 97, "undated": 0, "source": "table",
               "by_rule": { "numeric": 98 }, "beyond_reference": { "count": 0, "rules": {} } },
    "warnings": ["TEMPORAL_DUPLICATES_COLLAPSED/warn","TEMPORAL_DATES_TIED/info",
                 "TEMPORAL_WAVES_NEAR_DEGENERATE/warn","TEMPORAL_WAVE_SIGN_CONVENTION/info"],
    "primaeon": { "surface": "mcp-stdio", "seed": 42, "wave_sign": "canonical",
                  "score_invariable_sites": true, "scored_codons": 566, "taxon_cap": null,
                  "tree_free": null, "duplicates_collapsed": 1 }
  },
  "honesty": {
    "stage": "complete", "null_state": "finished", "calls_are_final": true,
    "uncalled_because": null, "wave_columns_pending": false,
    "permutations": { "requested": 1000, "completed": 1000, "cancelled": false, "skipped": false,
                      "grid_step": 0.000999, "q_min": 0.0839, "q_rank1_bound": 0.1678,
                      "work": 1.036e9, "budget": 5.0e10, "within": true,
                      "rng": "xoshiro256**", "seed": 42, "chunks": 10, "ms_per_draw": 1.167, "tested": true },
    "p_perm_fill": "p_perm and q_perm are 1.0 at the 398 codons that never reached stage two ... not a measurement (temporal.py:620-621). A 1.0 does NOT mean untested — a tested candidate every shuffle beat scores (1 + B) / (B + 1) = 1.0 exactly. The column that tells them apart is classification; on this surface the mask is get_results section=candidates.",
    "escape_hatch_used": false, "solitary_regime": false, "gate_vacuous": false,
    "wave_variance": {
      "shares_pct": [97.13, 1.48, 0.60, 0.47], "source": "confirmed-sweeps", "sign": "canonical",
      "conditioned_on": "the confirmed-sweep set, which is thresholded on the permutation p",
      "note": "these shares MOVE WITH THE NULL ... measured upstream on H1N1 at B = 100: 32 confirmed here against 18 there. Compare shapes and ordering, not digits."
    },
    "reference_command": {
      "command": "hyphaeon temporal -a alignment.fasta -t tree.nwk -d H5N1_HA_metadata.csv -B 1000 --time-points 250 --sweep-mode episodic --perm-alpha 0.05 --min-r2 0.35 -o temporal",
      "reproduces": false,
      "caveats": [
        "Running this does not make the four files diff clean ... 1 of 4,384 sites rows byte-identical on H1N1. Compare numerically, not with diff.",
        "The 1000 date shuffles this run drew came from xoshiro256** per-draw substreams (D17); hyphaeon temporal draws from numpy's MT19937 at a hard-coded RandomState(42). Both are valid nulls and neither is the other.",
        "Wave signs follow this application's canonical convention (D28); hyphaeon temporal has none."
      ]
    }
  },
  "next": "get_results job_id=... section=<summary|sites|curves|waves|permutations|dates|candidates|warnings|honesty|provenance>. The whole record is never returned inline: its trajectory store alone is megabytes."
}

get_results · 2 ms

{
  "job_id": "4368ab5e04a952f42fce650bcd9ff05c",
  "section": "sites",
  "top": 3
}

Result

{
  "analysis": "temporal", "section": "sites", "stage": "complete",
  "rows": [
    { "site": 123, "mutation_label": "R123S", "classification": "CONFIRMED_SWEEP", "cross_classification": "RESCUED_SWEEP",
      "is_confirmed_sweep": true, "is_rescued_sweep": true, "lrt": 0.128, "p_static": 0.3602, "q_static": 0.4032,
      "p_perm": 0.01099, "q_perm": 0.1319, "r2_fpca": 0.6784,
      "peak_date": 1998.458, "peak_intensity": 0.2021, "fwhm_years": 0.506, "auc": 0.1136, "Wave_1_loading": 15.797 },
    { "site": 288, "mutation_label": "G288S", "classification": "CONFIRMED_SWEEP", "cross_classification": "RESCUED_SWEEP",
      "is_confirmed_sweep": true, "is_rescued_sweep": true, "lrt": 1.081, "p_static": 0.1492, "q_static": 0.3285,
      "p_perm": 0.01099, "q_perm": 0.1319, "r2_fpca": 0.6784,
      "peak_date": 1998.458, "peak_intensity": 0.1776, "fwhm_years": 0.506, "auc": 0.0999, "Wave_1_loading": 15.797 },
    { "site": 455, "mutation_label": "D455Y", "classification": "CONFIRMED_SWEEP", "cross_classification": "RESCUED_SWEEP",
      "is_confirmed_sweep": true, "is_rescued_sweep": true, "lrt": 3.036, "p_static": 0.0407, "q_static": 0.2458,
      "p_perm": 0.01099, "q_perm": 0.1319, "r2_fpca": 0.6784,
      "peak_date": 1998.458, "peak_intensity": 0.1633, "fwhm_years": 0.506, "auc": 0.0916, "Wave_1_loading": 15.797 }
  ],
  "rows_returned": 3, "rows_available": 168, "codons_total": 566,
  "selection": "the 168 stage-one candidate codon(s), strongest peak intensity first (pass sites to name others; p_perm at a NON-candidate is the reference's assumed 1.0, not a measurement)",
  "truncated": { "rows": { "returned": 3, "total": 168, "ranked_by": "peak_intensity" } },
  "honesty": { "null_state": "finished", "calls_are_final": true, "uncalled_because": null }
}

Claude
The dates are clean, and that is worth saying before anything else: all 98 sequences matched the metadata table exactly, all 98 dated by its numeric date column, spanning 1996 to 2005, with nothing imputed and no metadata row left unmatched — so this analysis is about the dataset you submitted rather than a subset of it. One sequence was then collapsed as a duplicate haplotype, leaving 97 for the model. Over 566 codons, 168 passed the stage-one sweep-energy floor and 16 were confirmed as sweeps by the full 1,000-draw permutation null, which completed inside its work budget; all 16 are "rescued" rather than "concordant", meaning the static MEME-surrogate call did not flag them (only one codon reaches q_static ≤ 0.10) and the temporal signal is what identifies them. The strongest are R123S, G288S and D455Y, all peaking together in mid-1998 with a half-width of about six months — one wave, not three independent events, which the fPCA agrees with: the leading mode carries 97% of the variance and the run is flagged TEMPORAL_WAVES_NEAR_DEGENERATE, so treat the four modes as one shape plus noise rather than four dynamics. Three caveats I would not drop. The permutation q values bottom out at 0.13 because 168 candidates against 1,000 draws puts the BH floor there, so "confirmed" here means p_perm ≈ 0.011 rather than a controlled false-discovery rate. The wave variance shares are computed over the confirmed set, which is thresholded on a null drawn from a different generator than the reference's, so compare their shape and ordering with a command-line run, not their digits — the tool says so itself, and sets `reproduces: false` on the reproduction command for exactly that reason. And these are surrogate scores for MEME: the ranking is the result, and codons 123, 288 and 455 are the three to confirm with real MEME on Datamonkey before reporting them.

Reproducing a report

Every report page and every MCP result end with the same two things: the install line above, and the tool call that reproduces the run. The convention is fixed so a snippet can be pasted into Claude Code as it stands:

Files are file:///path/to/<name> placeholders carrying the file names the run had, never the sequences.
Option names are the CLI's flags in snake_case (model_variant, max_species, filter_p_thresh, n_permutations); a browser run's options are mapped to them one to one, and the presentation-only ones are dropped.
A run with no usable tree is use_tn93: true in the snippet, not a tree path: under D22 the app takes pairwise TN93 distances into the MDS, which is the reference's own --use-tn93.
The reference command that reproduces the same run with the Python package is in provenance.reference_command of every result.

claude mcp add hyphaeon -- npx @veg/hyphaeon-mcp

{
  "tool": "hyphaeon_analyze",
  "arguments": {
    "alignment": "file:///path/to/Smc6.fasta",
    "tree": "file:///path/to/Smc6.nwk",
    "model_variant": "general",
    "seed": 42
  }
}