Skip to content
Get packs

Guide · Agent skill

How to use the Adaptyv Protein Lab Assistant agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

The adaptyv skill is a set of instructions and helper code that lets an AI agent work with Adaptyv Bio Foundry, a cloud laboratory that measures protein binding, stability, expression and enzyme activity. This guide explains what the skill does, how to install it, how to connect it to your Foundry account, and how to make requests that stay within safe limits. By the end you will know how to ask the agent for a catalog search, a cost estimate, a batch draft and a results download, and where a human still needs to approve money being spent.

Adaptyv Protein Lab AssistantK-Dense-AI/scientific-agent-skills · MIT

Quick start

  1. 1Download and unzip the skill. Download the adaptyv ZIP from the item page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder. In the Claude apps, you can instead upload the ZIP under Settings, Capabilities, Skills.
  2. 2Create a Foundry token. In Adaptyv Foundry, go to Organization, Settings, Tokens and create a bearer token. Use Viewer for reads and Member for experiment writes.
  3. 3Set environment variables. Export ADAPTYV_API_KEY with your token. Optionally set ADAPTYV_TARGET_ID and ADAPTYV_EXPERIMENT_ID for the work you are doing.
  4. 4Test the connection. Run the curl command in the connect section. A list of targets confirms your token and network access are working.
  5. 5Install the SDK if you need scripts. Install adaptyv-sdk from the official GitHub repository with Python 3.11 or newer. Import from adaptyv, not adaptyv_sdk.
  6. 6Request a cost estimate. Ask the agent to build a spec and estimate the cost. Review the breakdown or warnings before anything is submitted.
  7. 7Approve the quote yourself. Confirm a quote only after a person with authority to spend has approved it, because confirming creates an invoice.

What the adaptyv skill makes the agent do

An agent skill is a folder of instructions (a SKILL.md file plus reference files) that an AI agent reads when a task matches the skill's description. The adaptyv skill covers the Adaptyv Bio Foundry API and its Python SDK. It applies to the Foundry target catalog, binding screening and affinity assays, thermostability, expression, fluorescence, epitope binning and enzyme activity workflows.

Once installed, the agent can search the target catalog, estimate experiment costs, create and submit a batch of protein sequences, track laboratory progress and retrieve results. The skill asks the agent to confirm the target construct, assay conditions, controls and replicate plan before a batch is submitted.

  • Search the target catalog for a named target
  • Estimate the cost of a protein experiment before submitting it
  • Create a draft experiment from a set of sequences
  • Submit a batch and retrieve its quote
  • Track lab progress and download results

How to install the adaptyv skill

Download the adaptyv ZIP from the item page and unzip it. For a personal setup, place the adaptyv folder inside ~/.claude/skills/. For a single project, place it inside that project's .claude/skills/ folder so it applies only there.

If you use the Claude apps instead, open Settings, go to Capabilities, then Skills, and upload the ZIP there. The skill folder should contain SKILL.md and references/api-endpoints.md.

  • Personal install: ~/.claude/skills/adaptyv/
  • Project install: .claude/skills/adaptyv/ inside your project
  • Claude apps: Settings → Capabilities → Skills, then upload the ZIP

Connect your Adaptyv Foundry account

You need an Adaptyv Foundry account. Create a bearer token in Foundry under Organization → Settings → Tokens. Use a Viewer token for reads and a Member token for experiment writes. Store it in the ADAPTYV_API_KEY environment variable. The optional variables ADAPTYV_TARGET_ID and ADAPTYV_EXPERIMENT_ID hold a target and an experiment you are working with.

The skill also documents FOUNDRY_API_TOKEN as an alternative name for the same token. If you keep settings in a project .env file, the code must call python-dotenv load_dotenv() itself, because the SDK does not load that file automatically. Never print or commit your token.

To check the connection, run the curl command below. A successful response lists targets. Resource endpoints need the bearer token, while the schema and the health endpoint are public.

curl --fail-with-body 'https://devs.adaptyvbio.com/api/v1/targets?limit=3' -H "Authorization: Bearer $ADAPTYV_API_KEY"

Install the Python SDK for scripts

If you want the agent to write or run Python, the skill expects Python 3.11 or newer and the adaptyv-sdk package installed from its official GitHub repository. The skill pins a specific reviewed revision of that repository. Import from the adaptyv package, not adaptyv_sdk.

The SDK's FoundryClient requires explicit api_key and base_url values. Direct REST examples use httpx.

  • Requires Python 3.11+
  • Install from the official GitHub repository, as shown in the command
  • Use FoundryClient with api_key and base_url set explicitly
uv pip install "git+https://github.com/adaptyvbio/adaptyv-sdk.git@cdf207819ed5a58e0c127453626d2bce125c8064"

Ask for what you need in plain language

The skill loads when your request matches its topic. You do not need to name the skill, but mentioning Adaptyv, Foundry, a target, or a binding or stability assay makes the match clear. Be specific about the target, the assay method and the sequences you have.

Good requests state the goal and the constraints. For example, ask the agent to find EGFR targets in the catalog and list their vendors, or to estimate the cost of a BLI screening batch of your candidate file with three replicates before anything is submitted.

  • Find targets in the Adaptyv catalog that match a protein name, and list vendor and catalog number
  • Estimate the cost of a screening batch using candidates.json and three replicates
  • Check whether my affinity experiment has the antigen concentrations it needs
  • Download the results for experiment ID from my environment and save them as JSON
  • Explain what the quote says and what it will cost before I approve it

Estimate costs and submit safely

The agent builds a spec with an experiment type, method, target, sequences and replicate count, then asks for a cost estimate. If the estimate comes back without a full breakdown, the skill says to review the warnings and obtain a full quote rather than proceed.

Submitting an experiment and asking for a quote are separate from paying for it. Confirming a quote creates an invoice, so the skill tells agents not to confirm a quote without explicit human authorization. Keep that approval step with a person who is allowed to commit funds. Quote generation is asynchronous, so a 404 right after submission can be normal; retry with a bounded timeout.

  • Review the target, construct, assay conditions, controls and replicates first
  • Treat an incomplete estimate as a reason to stop and get a full quote
  • Only confirm a quote after a human has authorized that spend

Retrieve and archive results

When an experiment is finished, the skill recommends using REST JSON to archive results, because the SDK's generated models may drop some newer fields. The example checks that results are complete, pages through them, and writes them to foundry-results.json.

Archive raw data packages when available, with sequence and target identities, assay method, units, conditions, replicate measurements and fit quality. Null kinetic values mean a missing or unresolved measurement, not zero. A screening binding call is not a measured affinity, so compare KD values only under compatible conditions.

  • Wait until the experiment results status is complete before analysis
  • Keep units, conditions and replicate data with each value
  • Do not treat null values as zero or a screening call as an affinity

Safety notes for scripts and network access

This skill runs scripts and uses the network. Its scripts contact devs.adaptyvbio.com, and the token grants access to your organization's Foundry data and the ability to create invoices. Read any script the agent writes before you run it, keep the token in an environment variable or ignored .env file, and never paste it into a chat or commit it to a repository.

The skill is licensed under MIT. Its guidance targets Adaptyv's API, catalog and SDK, so it is not a general lab guide for other services.

  • Network access to devs.adaptyvbio.com is required
  • Never print, log or commit the bearer token
  • Review generated scripts before running them against a paid account

What people use it for

  • Screening a list of candidate protein sequences for binding to a catalog target using a BLI screening batch
  • Checking an affinity experiment spec, including antigen concentrations, before submitting it for a quote
  • Scripting a repeatable pipeline that submits batches and archives full results as JSON
  • Tracking the progress of a running experiment and downloading its results once they are complete
  • Comparing target vendors and catalog numbers before choosing a target ID

Frequently asked questions

How do I install the adaptyv agent skill?

Download the ZIP from the item page and unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder for one project. In the Claude apps, upload the ZIP under Settings, Capabilities, Skills instead.

What do I need to use the adaptyv skill?

You need an Adaptyv Foundry account with a bearer token (Viewer or Member role), the ADAPTYV_API_KEY environment variable, and network access to devs.adaptyvbio.com. Python scripts need Python 3.11 or newer and adaptyv-sdk, and direct REST examples use httpx.

Is the adaptyv skill free?

The skill is licensed under MIT. Using the Adaptyv lab service involves a Foundry account and money is involved in quotes and invoices, so the skill cannot run experiments without a paid account.

Why does my adaptyv get_quote call return 404?

Quote generation is asynchronous, so a 404 soon after a successful submission can be normal. Retry with a bounded timeout. A draft experiment has no quote yet.

Can the agent buy experiments automatically?

The skill warns that confirming a quote creates an invoice, so agents should not confirm quotes without explicit human authorization. Keep that approval step with a person.

Are there alternatives to the adaptyv skill for protein lab work?

This skill is specific to Adaptyv's API, catalog and SDK, so it does not help with other labs or services. For a different lab, you would need that provider's own tools or documentation.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Use this skill if you send protein candidates to Adaptyv Bio Foundry for binding, stability or expression screening, or if you build scripts against its API or SDK and want an agent to check specs and costs first. Skip it if you do not have a paid Foundry account, if you want an agent to buy experiments without a human approving the quote, or if you work with a different lab service.

Download the skill