Query Tahoe's Frontier Scale Datasets from your AI agent.

Tahoe's Frontier Scale Datasets hold single-cell perturbation data across drugs and cell lines. Ask in plain language for expression and differential expression, pathway scores, phenotype readouts, embedding distances, perturbation markers and drug and cell-line annotations, or run guarded read-only SQL — no downloads, just one install and a sign‑in.

A short tour: install, ask, explore.

Example questions

You Which MEK and RAF inhibitors are in Tahoe's data?

Tara 4 MEK (Binimetinib, Cobimetinib, TAK-733, Trametinib) and 3 RAF (Dabrafenib, Encorafenib, Vemurafenib).

MEKBinimetinibCobimetinibTAK-733TrametinibRAFDabrafenibEncorafenibVemurafenib

Tools used: list_releases, get_drug_info

Does binimetinib up- or down-regulate KRAS signaling?
Which MOA most raises the interferon α/β response at the highest dose?
Is pitavastatin's stronger suppression of YAP signaling than pravastatin driven by greater on-target activity?
Which genes does trametinib change most versus vehicle in a melanoma cell line, and is DUSP6 among the top down-regulated?
Which drugs most deplete skin (melanoma) lines relative to the other pooled lines, and do they also move cells furthest from vehicle?

Install

Paste this into your agent:

Install the Tahoe plugin: add the marketplace tahoebio/tahoe-public-plugins, install the tara plugin, then sign me in.

In Claude Code:

claude plugin marketplace add tahoebio/tahoe-public-plugins
claude plugin install tara@tahoe-public

In Codex:

codex plugin marketplace add tahoebio/tahoe-public-plugins
codex plugin add tara@tahoe-public

Connect the server directly, without the plugin:

claude mcp add --transport http --scope user tahoe-public https://tara.tahoebio.io/mcp

codex mcp add tahoe-public --url https://tara.tahoebio.io/mcp

You'll be asked to sign in. If tools don't appear, restart your session. Preview access: invited accounts only.

What you get

Find & resolve

Pin a release, then find the drugs, doses, cell lines and genes to ask about.

Which MEK and RAF inhibitors are in Tahoe's data? 4 MEK (Binimetinib, Cobimetinib, TAK-733, Trametinib) and 3 RAF (Dabrafenib, Encorafenib, Vemurafenib).

list_releasesThe dataset release this service serves. Call this first.
search_conditionsFind conditions — drug, dose and cell line — in the index.
resolve_genesResolve gene symbols and Ensembl ids; exact, ambiguous or not_found, never a guess.
list_cell_linesThe cell lines: Cellosaurus id, name, DepMap id, organ and driver annotations.
get_drug_infoDrug annotations: targets, mechanism class, approval status; filter by drug, moa or target.

Expression

Gene expression for drugs × cell lines, against matched vehicle controls.

Which genes does trametinib change most versus vehicle in a melanoma cell line, and is DUSP6 among the top down-regulated? DUSP6 is down-regulated by trametinib in C32 at 0.5 µM on both plates (log2 fold change −5.24 and −0.95), though not among the top genes returned by adjusted p-value.

get_expression_slicePseudobulk expression for genes × conditions, with matched controls and the release and request it answered.
get_differential_expressionDrug-versus-vehicle differential expression per contrast, with the release and request it answered.
get_perturbation_markersThe genes a classifier found most important for one drug × cell line condition.

Pathways

Pathway-level (VISION) scores in place of single genes.

Does binimetinib up- or down-regulate KRAS signaling? Down-regulates: the median Hallmark KRAS score is negative at every dose.

search_gene_setsFind gene sets (pathways) by name or collection, to pass to get_pathway_scores.
get_pathway_scoresDifferential VISION pathway scores per drug × cell line, with the release and request it answered.

Phenotypes & embeddings

Whole-cell readouts and how far a drug moves a cell line's state.

Which drugs most deplete skin (melanoma) lines relative to the other pooled lines, and do they also move cells furthest from vehicle? Epirubicin in SK-MEL-2 at 5 µM has a relative-depletion score of −2.61 (not absolute viability); among these sampled conditions, depletion and embedding-distance rankings differ.

get_phenotype_scoresSurvival, cell-cycle, heterogeneity and separability readouts for drugs × cell lines.
get_embedding_distanceHow far a drug moves a cell line's state from vehicle, in the release's embedding space.

SQL & guidance

Your own read-only queries, the analysis guide, and a liveness check.

Which MOA most raises the interferon α/β response at the highest dose? MEK inhibitors (Hallmark interferon-alpha response, 5 µM).

run_sqlRead-only SQL over the approved tables, guarded and cost-bounded; the exact SQL is returned.
get_guideHow to analyse and interpret the data: controls, units, coverage traps.
healthLiveness, the served release, and the response contract version.

Every answer shows its work

Completed analysis results name the data release and record the request they answered.
SQL is read-only and checked before it runs; the exact SQL is returned so you can inspect and re-run it.
If a result was trimmed to fit, it says so, names the reason and states how many rows it returned.
Sign in required; data is CC0 1.0.

Free during preview — fair-use limits

30 queries per hour per user (run_sql counts as 3; fetching a pending result counts as 1).
Results arrive in seconds, or as a handle to fetch later (kept about 12 hours).
Each result holds up to 1,000 rows and 40 KB.
Preview access: invited accounts only.

FAQ

Who can use it?

During the preview, invited accounts only. Sign in with the account you were invited with; ask the Tahoe team for access.

What data is in it?

Tahoe's Frontier Scale Datasets — single-cell perturbation data across drugs and cell lines; not DepMap or PRISM. DepMap ids appear only as cell-line labels. More datasets are planned.

What is the licence?

What are the limits?

30 queries per hour per user, up to 1,000 rows and 40 KB per result. See the fair-use limits above.

What if the tools don't appear?

Restart your session after signing in.