Your AI agent already speaks Unfold.
Create structured plans, distribute them via claim links, and track progress -- all from your AI agent. Four lines of config.
{
"mcpServers": {
"unfold": {
"command": "npx",
"args": ["@anthropic/unfold-mcp-server"],
"env": { "UNFOLD_API_KEY": "unfold_sk_..." }
}
}
}Works with Claude Code, Cursor, Windsurf, and any MCP client.
What your AI agent can do
Assess skills. Close gaps. Track growth.
Available tools
Eleven tools. Assess skills, create targeted learning paths, track individual progress, and analyze cohorts.
create_goalCreate a goal with AI-generated plan. Agent auto-answers clarification questions using context you provide. Tag with metadata for cohort analytics.
title, description, context?, auto_respond?, priority, metadata?goal_id, claim_link, progress_link, questions?, agent_answers_used?get_goal_statusCheck progress: completion %, step-by-step detail, time spent, blockers, who claimed it, and custom metadata tags.
goal_idprogress, steps[], metadata?, claim_created_at, assigned_toget_analyticsAggregated cohort analytics: completion rates, claim activation, at-risk learners, step-level funnel, and resource engagement. Filter by metadata, date range, or group by any dimension.
group_by?, inactive_days?, metadata?, date_from?, date_to?, include_funnel?completion_rate, at_risk_goals[], step_funnel[], completion_by_dimension[]list_goalsList goals in your org. Filter by status, claim state, metadata tags, assigned learner email, or inactivity window for at-risk detection.
status?, claim_status?, metadata?, assigned_email?, inactive_days?goals[], metadata, claim_created_at, total_countget_clarificationGet pending clarification questions with agent-suggested answers and confidence levels. Use after create_goal with auto_respond=false.
goal_idquestions[], agent_answer, agent_confidence, agent_sourcesubmit_clarificationSubmit your answers to clarification questions. Accept agent suggestions for the rest, or override any answer.
goal_id, answers?, accept_agent_answers?goal_id, status, agent_answers_used[]import_planImport pre-formulated steps from Jira, Linear, or your own tool. AI enriches with dependencies, durations, severity, and critical path.
title, steps[], enrich?, enrich_options?, metadata?goal_id, plan_id, enriched_steps[], claim_linkrevoke_claimInvalidate a claim link. Already-claimed goals are not affected.
claim_tokensuccess, messagegenerate_skill_assessmentGenerate AI-validated MCQs for a skill and proficiency level, anchored to a work item. Returns a signed token for tamper-proof scoring.
work_item_context, skill, target_proficiency, num_questions, difficulty_mix?, request_idassessment_token, questions[], band_map, max_raw_score, model_metascore_skill_assessmentScore answers against the signed token. Returns proficiency band, gap vs target, and a suggested goal seed when the learner falls short.
assessment_token, answers[], band_thresholds?, request_idraw_pct, band, gap_bands, per_question[], recommended_action, suggested_goal_seed?get_assessment_capabilitiesIntrospect supported parameters before generating: languages, question range, band thresholds, difficulty mix defaults.
(none)schema_version, supported_languages, min/max_questions, default_band_thresholdsWhat you say to your AI agent
No code to write. Just describe what you need.
“Create an onboarding plan for our new hire starting Monday and tag it with department=engineering and cohort=q2-2026”
Team onboarding with metadata“How is the spring 2026 frontend track doing? Show me completion rates and where learners are dropping off”
Cohort analytics“Which students haven't made any progress in the last 7 days? I want to reach out to them”
At-risk detection“Import our Jira sprint backlog as a goal with dependencies and time estimates”
Plan import + enrichment“Create a Python learning path but let me review the clarification questions before generating the plan”
Semi-auto with review“Compare completion rates across all our training tracks this quarter”
Program comparison“Generate a Python assessment with 8 questions for our ML engineers who need medium proficiency to work on the feature pipeline”
Skill assessment generation“Score this assessment and if the learner didn't reach the target, create a learning path from the results focusing on their weak areas”
Assess then learn flow“Before assigning the new project, assess the team on SQL and create personalized upskilling goals for anyone below medium”
Team skill gap analysisPrefer REST? We have that too.
All MCP tools map to REST endpoints. Ten endpoints covering goal creation, skill assessments, individual progress, cohort analytics, and claim management. Authenticate with your API key as a Bearer token.
POST/api/v1/ext/goals/unfoldCreate goal with agent-assisted plan generationPOST/api/v1/ext/goals/importImport steps with AI enrichmentPOST/api/v1/ext/goals/:id/clarify/submit-allSubmit clarification answersGET/api/v1/ext/goals/:idGet goal status, step detail, metadata & agent answersGET/api/v1/ext/goalsList goals -- filter by status, metadata tags, assigned email, inactivityGET/api/v1/ext/analyticsCohort KPIs, at-risk learners, step funnel, completion by dimensionDELETE/api/v1/ext/goals/claims/:tokenRevoke a claim linkPOST/api/v1/ext/assessments/generateGenerate skill assessment MCQs with signed tokenPOST/api/v1/ext/assessments/scoreScore answers against signed token, get proficiency band + gapGET/api/v1/ext/assessments/capabilitiesIntrospect supported skills, question range, band defaultscurl -X POST https://api.unfoldit.ai/api/v1/ext/goals/unfold \
-H "Authorization: Bearer unfold_sk_..." \
-H "Content-Type: application/json" \
-d '{
"title": "Python Certification",
"autoRespond": true,
"context": { "experienceLevel": "beginner", "timeline": "12 weeks" }
}'