Connecting AI tools and installing skills
Connect the Nadoo Bio MCP server to AI tools such as Claude Code, Codex, and Cursor, and install agent skills.
The Nadoo Bio MCP server is at https://nadoo.bio/mcp and uses streamable HTTP. Send the API key in the Authorization: Bearer header. MCP OAuth is not yet available, so use a client that accepts headers. Browser login and MCP API-key authentication are separate.
- Prepare API key
- Connect AI tool
- Check connection
- Install skills
1. Prepare an API key
- Open API keys in the left menu and choose a name, scopes, and expiry (30, 90, or 365 days; default 90). Copy the key immediately: it is shown only once after creation.
- Store the key in the
NADOO_BIO_API_KEYenvironment variable. The command below keeps it out of the screen and shell history. To use it in a new terminal, configure your shell or a secrets manager. Later commands read this variable, so you do not paste the key into the command line.
read -rs NADOO_BIO_API_KEY && export NADOO_BIO_API_KEY # Paste the key and press Enterlist_tools, get_tool_schema, search_docs, and read_doc work without a key. Other tools require a key with the scopes below. To let an AI tool create and submit plans, grant both plan creation and job submission scopes. Plans created by AI still require a person's approval on the web.
| Scope | UI label | MCP tools requiring this scope |
|---|---|---|
catalog:read | Read tool catalog | validate_input |
jobs:read | Read jobs and results | get_candidate_selection, preview_candidate_selection, list_candidate_inputs, inspect_input, list_projects, list_artifacts, list_jobs, get_job, get_result, query_result_table, read_result_file, preview_structure_comparison, create_structure_comparison, list_structure_comparisons, get_structure_comparison, list_result_views, render_result_view, get_job_events, list_workflow_runs, get_workflow_run, get_plan, list_plans, get_provenance, create_view_link |
plans:write | Create execution plans | list_execution_options, prepare_design_input, create_plan, label_plan, rerun_plan, create_batch_plans, preview_batch_plans |
jobs:submit | Submit jobs | submit_plan, submit_delegated_plan, submit_batch, cancel_job |
files:write | Upload files | create_upload, get_upload, complete_upload, cancel_upload, create_structure_comparison |
2. Choose a connection method
Choose the instructions for your AI tool below. After connecting the MCP server, install the skills in section 5 to add guidance for Nadoo Bio workflows.
3. Manual client setup
Claude Code
claude mcp add --transport http nadoo-bio -s user https://nadoo.bio/mcp --header 'Authorization: Bearer ${NADOO_BIO_API_KEY}'-s user registers the server at user scope for all projects. Without it, registration is local to the current project. Inside single quotes, the shell leaves ${NADOO_BIO_API_KEY} unexpanded, so it is stored literally and Claude Code replaces it with the environment variable when it connects. The key value is therefore never written to the configuration file, and NADOO_BIO_API_KEY must be set in the environment that launches Claude Code. Do not use double quotes: the shell would expand the key and Claude Code would store it as is.
Codex
codex mcp add nadoo-bio --url https://nadoo.bio/mcp --bearer-token-env-var NADOO_BIO_API_KEYCodex stores only the environment variable name and reads its value at startup. NADOO_BIO_API_KEY must be available in the shell or app environment that launches Codex.
Clients using JSON configuration
| Tool | Configuration file | URL key |
|---|---|---|
| Cursor | .cursor/mcp.json (project) or ~/.cursor/mcp.json | url |
| Windsurf | ~/.codeium/windsurf/mcp_config.json | serverUrl |
| Gemini CLI | ~/.gemini/settings.json | httpUrl |
| Cline | cline_mcp_settings.json (open from Cline's MCP settings) | url; also add "disabled": false |
This example is for Cursor. For other clients, change the URL key to match the table.
{
"mcpServers": {
"nadoo-bio": {
"url": "https://nadoo.bio/mcp",
"headers": {
"Authorization": "Bearer <API_KEY>"
}
}
}
}This example stores the API key directly in the configuration file, so do not commit it to a repository. To use environment variables or secret inputs, follow the configuration syntax supported by your AI tool.
VS Code
In the command palette, choose MCP: Add Server, select HTTP, and enter the address https://nadoo.bio/mcp and name nadoo-bio. Add the Authorization header to the resulting server configuration. See the VS Code documentation for headers and secret inputs.
Connectors that accept only a URL
To use MCP tools that require an API key, the connection must support authentication headers. If a connector accepts only a server URL, use another connection method that supports headers. Nadoo Bio does not provide OAuth for MCP connections, and web sign-in cannot replace API key authentication.
4. Check the connection
First, use get_connection_context to check the account, organization, and scopes of the API key. A successful public listing does not confirm authentication. If legacyUnscoped is true, the key uses legacy role permissions rather than minimum scopes. This check does not grant project access or approve paid execution. Web links check the signed-in browser account separately; never put the key in a URL.
- Claude Code:
claude mcp list, or/mcpin a conversation. Codex:codex mcp list. - Ask your AI tool: "Show me the Nadoo Bio tool list."
list_toolsworks without a key, so if it fails, check connection settings before authentication.
Hand off a paid job approved on the web
For USD LIVE plans supported by the active deployment, open Prepare agent delegation on the approval page, select the key, and record review and consent. Use Create submission request → Copy submission request in the delegation details, send the JSON to the agent, and ask it to submit. Opening or copying a link does not run a job. Buttons depend on the current approval session, deadline, and execution support. Signing in again alone does not restore the original approval authority.
The agent uses the copied submit_delegated_plan arguments unchanged and preserves the original requestKey and body. Recover a lost response with the same command. The designated API key and jobs:submit are required; querying get_job after acceptance separately requires jobs:read. Without read permission, open the returned job link rather than treating acceptance as failed or submitting again. Keep the API key in connection settings, never in commands or URLs. Delegation is not supported for every model or currency path.
5. Install agent skills
Skills are SKILL.md instructions that help AI tools use Nadoo Bio MCP tools in the right order. Skills do not provide the tools themselves; connect MCP first.
| Skill | When to use it | First 12 sha256 characters |
|---|---|---|
nadoobio-account-and-approval | Use when permissions, approval, or usage limits block progress. Explains errors such as scope_required, not_found, plan_not_approved, monthly_limit_exceeded, and context_changed, and what to ask the user. Prevents bypasses or access to guessed resources. | a6dd78911a1e |
nadoobio-batch-screening | Use to prepare plans for multiple candidates and check whether they can be submitted. Calls create_batch_plans with stored CSV/FASTA data instead of repeating single-item calls. Follow approval.status and reason for approval and submission availability. | 8689bbba1619 |
nadoobio-complex-analysis | Review chain correspondence, contacts, and arrangement metrics in structures from complex or targeted protein-degradation research. | 6f6312f2a21c |
nadoobio-docking-screening | Assess input preparation and score comparability for molecular docking and candidate screening. | 000ad86c87dd |
nadoobio-dynamics | Check trajectories, time, periodic boundaries, and interpretation in molecular-dynamics simulations. | 6a981f354bbc |
nadoobio-folding | Check input requirements and the meaning of structural confidence when predicting structures from protein sequences or reviewing folding results. | b4c914327bd5 |
nadoobio-input-preparation | Use this skill to select protein research tools and prepare and validate inputs in Nadoo Bio. Discover tools with list_tools, read get_tool_schema, and use validate_input and inspect_input to check sequences and stored files against the contract. Check supported design-region preparation. This precedes execution-plan creation. | a6af4a9d52f7 |
nadoobio-metric-interpretation | Explain evidence levels without confusing structural confidence, comparison error, candidate scores, and experimental measurements. | 366163d78bd8 |
nadoobio-plan-and-submit | Use to run jobs in the order list_execution_options → create_plan → human approval → submit_plan → get_job polling. Covers approval links, idempotent retries, and monthly limits. Prevents execution without human approval or retrying the same request under new keys. | c84e012f5fdd |
nadoobio-protein-design | Define inputs for antibody/protein sequence or structure design and distinguish design from validation. | 186004f4de2f |
nadoobio-reproducibility | Use to reproduce or cite earlier results, or rerun the same inputs. Covers planHash, release, and policy in get_plan; lineage in get_provenance; and the source-plan link in rerun_plan. Fixed inputs do not guarantee identical outputs. | 4445c2b467eb |
nadoobio-research-routing | Map research intent to structure prediction, docking, design, complex analysis, trajectories, or metric interpretation and find currently executable tools. | fe09e436a89d |
nadoobio-result-interpretation | Use to read job results and choose next steps. Check verified artifacts with get_result and read small text results such as metric CSVs with read_result_file. Explains metrics without confusing predictions, structural confidence, actual binding, and experimental validation. Also compares structure files or retrieves saved comparisons with numbers, images, and links. Give the user the results-page link. | e8318245b12f |
nadoobio-structure-comparison | Use to compare two protein structures. Select models, chains, and regions from verified job results or stored files, then calculate, save, and reopen superposed structures, Cα RMSD, and per-position deviations. | d911434bfe0f |
The open-source skills CLI can install directly from the Nadoo Bio documentation domain. It finds skill folders for Claude Code, Codex, Cursor, Copilot, Windsurf, Cline, Gemini CLI, and other clients. Update later with npx skills update.
npx skills add https://nadoo.bio # Choose clients and skills
npx skills add https://nadoo.bio --all -g # Install all skills at user scopeThe domain's /.well-known/agent-skills/index.json provides the skill list and sha256 values. The skills CLI uses them to verify files.
The MCP server also provides the same skills through the Skills extension (io.modelcontextprotocol/skills) and resources/list. If your AI tool cannot read MCP skill resources, use the file installation method above.
Skills are available under the MIT license and may be copied, modified, and redistributed. Use of Nadoo Bio itself is governed by its terms of service.
6. Rotate keys and remove connections
Before a key expires, issue a new key and update NADOO_BIO_API_KEY. Codex and Claude Code registered as in section 3 only need the variable updated; JSON configuration files store the key value, so edit those files. If you registered Claude Code with double quotes as in earlier instructions, the key value is still in its configuration file: remove the server once and register it again with the command in section 3. When the rotation is complete, revoke the old key on the API keys page.
claude mcp remove nadoo-bio -s user
codex mcp remove nadoo-bioTroubleshooting
| Symptom | What to check |
|---|---|
invalid_credentials or authentication_required | Check for typos, expiry, or revocation. Suspending or removing a member also blocks that person's keys. |
scope_required | The error's required value names the missing scope. Issue a new key with that scope. |
agent_execution_disabled | AI agent execution is disabled in service settings, so only reads are available. Contact your administrator. |
| Need to connect to another server | In your AI tool settings, change the server URL to the target server’s /mcp endpoint. |