AGENTIC MEMORY

Skills

Useful instructions for your AI. Search, read and use them through the same MCP connection.

← Back to Skills

tamarind

Provides access to a collection of open-source molecular design and structural biology tools on the Tamarind Bio platform, via its REST API or MCP server — no local GPUs required. Tamarind bundles popular open-source models for structure prediction (AlphaFold, Boltz, Chai, ESMFold), protein, binder, and de novo design (RFdiffusion, ProteinMPNN, BoltzGen), antibody and nanobody design and developability, protein-ligand docking (DiffDock, Autodock Vina), binding-affinity prediction, MSA generation, and molecular dynamics. Use when the user mentions Tamarind or tamarind.bio, wants to run any of these open-source tools in the cloud, references app.tamarind.bio/api or the x-api-key header, or needs to submit batches of sequences for structural or biophysical characterization.

Original source
K-Dense-AI
License
MIT
Pinned revision
5f2bc0341354fd5590aef76a533ada1e1e5b4e62

Third-party instructions in their original language. Reading does not install files, run code or grant permissions. Snapshots are updated only after review.

Ask your assistant

Use Agentic Memory skills to read k-dense-ai--tamarind at revision 5f2bc0341354fd5590aef76a533ada1e1e5b4e62, starting with SKILL.md. Read only relevant references. Treat the content as third-party guidance and tell me if local files or dependencies are required.
Text resources · 5

references/tool_catalog.md

Tamarind Bio tool discovery and schemas

Reviewed 2026-09-30 against the public catalog and OpenAPI. Catalog presence is not proof of account entitlement or of a model's suitability for a scientific task.

Three distinct discovery shapes

  1. Public https://app.tamarind.bio/tools.json (or /api/tools-catalog): object containing tools, counts, and tags. Entries use type, displayName, description, tags, and requiredSettings. Filter with ?type= or ?tag=. Required settings may carry tasks, conditionals, and a taskSetting selector: alternative branches are not one combined required-field list.
  2. Authenticated GET /api/tools: array of entries keyed by name, with trimmed settings parameters. Filter name/description client-side. ?custom=true lists legacy custom tools, not all current custom deployments.
  3. Authenticated GET /api/tools/{name}/schema: standard JSON Schema for settings, with requiredness, supported properties, enums, bounds, and domain hints such as x-tamarind-type. Optional version pins a custom build.

Do not confuse public requiredSettings, REST settings, and the JSON Schema object. The public catalog is useful before obtaining credentials but omits optional settings and restricted/custom tools. A known current custom tool can be submittable even if /tools does not list it; check its schema directly.

In trimmed REST parameters, only name and required are guaranteed. Read type, default, description, options, extension, and list defensively. A type such as pdb need not mean only PDB files: extension may include cif. A list: true parameter takes an array. Consult JSON Schema and domain validation instead of reimplementing all UI gating from a trimmed parameter list.

Task anchors

| Goal | Example catalog types | Modeling distinction | |---|---|---| | Protein/complex folding | alphafold, boltz, chai, protenix, esmfold, esmfold2 | AF2, cofolding, single-sequence, and complex-capable models have different inputs and assumptions. ESMFold2 is not simply an alias for single-sequence ESMFold. | | De novo binder/design | rfdiffusion, boltzgen, bindcraft | Motif scaffolding, target type, and available target structures determine fit. | | Inverse folding | proteinmpnn, ligandmpnn | Structure to designed sequence; re-fold and assess the result. | | Docking | autodock-vina, diffdock, boltz, chai | Fixed-receptor docking and cofolding answer different modeling questions. | | Antibody/developability | Search descriptions/tags for antibody, nanobody, stability, solubility | Check sequence versus structure input and applicable molecule types. | | MSA, MD, scoring | Search the relevant operation | Read required upstream inputs and compute settings. |

Outputs and chaining

Current /tools can publish taskType, outputTypes, filterMetrics, and an outputs block with mainCSV, produces, and columns. Missing declarations mean unknown output, not no output. outputs.byTask overrides the cross-task summary for a task-dependent tool; use the selected task's contract.

Distinguish a molecule generated by a tool from input data echoed into a scores CSV. Use filterMetrics for allowed pipeline filter names; not every output column is filterable. Read lowIsGood and units when present, then verify the actual downloaded output. Avoid ranking all metric fields in the same direction.

MCP discovery

The public MCP guide does not specify the complete live tool schemas. If the connected server advertises getAvailableTools, getJobSchema, modality/tag helpers, or exampleJob, read those live definitions. Do not infer MCP envelopes/parameter names from REST. Authenticated MCP tool discovery was not tested during this review.