Power your agent
The publisher's own URL. Not a PlugMCP endpoint.
Host a copy with our keys and logs.
https://agent-tune.com/mcpThe Claude button above opens Claude with the name and URL filled in. You still confirm it there.
claude mcp add --transport http com.agent-tune-agenttune https://agent-tune.com/mcpRun in a terminal.
{ "mcpServers": { "com.agent-tune-agenttune": { "command": "npx", "args": [ "-y", "mcp-remote", "https://agent-tune.com/mcp" ] } }}claude_desktop_config.json
Remote URL or a command on your machine? How the two installs differ
Run on PlugMCP
A new server you create. It is not this listing's official endpoint. Use the publisher install when you want their URL and their credentials.
LaunchFrom the publisher
Editable communication preferences for AI assistants. Browse 43 templates, choose the rules that fit, and review them before adding them to your instructions. A personality label is optional; templates are editorial suggestions, not diagnoses or proven performance improvements.
Website · Preference builder · ·
Personality tuning files for AI agents: 43 MIT-licensed tunings + 5 inline personality tests.
Reading a resource does not authorize installation, an external action or a lasting change to preferences. Templates must preserve accuracy, material uncertainty and authorized scope.
Combine only rules you want. When two rules conflict, choose explicitly rather than treating a score or framework as an instruction priority. The framework labels are navigation aids, not claims about every person who uses a label.
The current release includes five interactive questionnaire adaptations: MBTI-style, Enneagram, DISC, attachment and IPIP Big Five. The Big Five adaptation returns raw dimension totals and item means. It does not return population percentiles or select instructions automatically. Review the availability and rights record for other instruments. All five questionnaires are available. Their source terms remain separate; availability does not establish unrestricted reuse rights.
See third-party notices before reusing instrument material. Original AgentTune code and preference templates are MIT licensed; that does not make every third-party instrument MIT or commercially reusable.
This repository owns the canonical content. The website consumes a pinned release, not a runtime fetch from main. dist/manifest.json gives the release version and SHA-256 of every exported artifact. dist/instruments.json identifies the available definitions, item IDs, response anchors and definition hashes. Every scoring result identifies its instrument and scorer version.
const { score, scoreOrdered } = require('./dist/score.js');
const instrument = require('./dist/instruments.json').find(d => d.route === 'big-five');
const result = score({
instrumentId: instrument.id,
instrumentVersion: instrument.version,
responses: instrument.items.map(item => ({ itemId: item.id, value: 3 }))
});
// result.status === 'complete'; all totals 30 and all item means 3.
// scoreOrdered(id, version, answers) requires the exact complete item order.
Valid partial responses return incomplete, with missing item IDs and no final score. Duplicate/unknown IDs, noninteger/out-of-range values and malformed responses return invalid. Missing answers are never silently imputed. Raw scores are separate from a user's subsequent choice of communication preferences. Keep personal responses local and out of public pull requests.
Browser: load dist/score.js to use AgentTuneScoring. It contains scoring keys for the approved instrument without embedding question wording. The original generic engine is also available separately as dist/engine.js; users are responsible for rights to any definitions they register.
The site exposes body-only Markdown at /resources/tunings/<system>/<slug>.md, metadata-rich mirrors via the tuning catalog, and a versioned content manifest. Apply only the body; metadata is for discovery and verification.
The stateless MCP endpoint is https://agent-tune.com/mcp. Its tools are list_tunings, get_tuning, get_test_spec, list_resources, get_resource and get_free_tools. Questionnaire availability is explicit; an unavailable specification must not be treated as an instruction to administer an old questionnaire. MCP retrieves public resources and does not need your questionnaire answers.
Use the installation protocol and platform registry for current destinations. Metadata separates verify.saved_text from verify.behavior; there is no universal compliance probe.
Run npm run build and npm test with Node 22 or later. No dependency installation is needed. Edit data/tunings.json, data/contract.md and approved definitions in data/instruments/; do not edit generated copies. npm run check also detects stale generated artifacts. License eligibility, instrument IDs, fixture scores and export parity are release gates.
See the changelog and legacy provenance. Historical questionnaires contained adaptations and undocumented substitutions. Their responses must remain attached to their actual versions. Correcting future instruments does not retroactively correct previous observations. Research about model questionnaire responses is distinct from evidence that a tuning improves task performance.
MIT for original AgentTune work, with the third-party exceptions and availability policy. Review the per-instrument terms and provenance before redistribution.
© 2026 MCP Directory.
Personality tuning files for AI agents: 43 MIT-licensed tunings + 5 inline personality tests.
{ "mcpServers": { "com.agent-tune-agenttune": { "url": "https://agent-tune.com/mcp", "headers": {} } }}Tools
No tool list yet. Remote servers get one after the first health check.
FAQ
Copy the http config for Cursor, Claude Desktop, Claude Code, or VS Code on this page. It uses the publisher's install, not a PlugMCP proxy.
No required environment variable is listed on this record.
Agenttune is reached over HTTP.
Last checked 2026-11-08. Personality tuning files for AI agents: 43 MIT-licensed tunings + 5 inline personality tests.