CueBench for Developers is live: score how well you drive coding agents(anthropic.com)

{{ dashKicker }}
{{ dashTitle }}

Updated {{ updatedAt }}

{{ ac.label }}
{{ ac.value }}
{{ ac.sub }}

{{ heroLabel }}
{{ heroBody }}

{{ v.label }}
{{ v.delta }}

{{ v.score }}

{{ v.spark }}

Cost & efficiency

{{ costTotal }}
Total spend

{{ costAvg }}
Avg per session

Model usage

{{ mr.model }}

{{ mr.count }}

No session data yet — cost metrics will appear once sessions are recorded.

Risk signals

{{ riskLoopsDisplay }}

Sessions with loops
of total sessions flagged

{{ riskBelowThreshold }}

Below threshold
developers below {{ alertBelow }}-point target

Review developers →

{{ c.tag }}

{{ c.initials }}

{{ c.name }}
{{ c.role }}

{{ c.value }}
{{ c.valueSub }}

People & Skills
Developers

Updated {{ updatedAt }}

! ALERT
{{ alertText }}
Review {{ alertCount }} developers

Side-by-side comparison
Clear ×

{{ ce.name }}
{{ ce.role }} · {{ ce.dept }}

{{ cv.label }}

{{ cv.score }}

Overall
{{ ce.overall }}

{{ tableTitle }}
{{ tableMeta }}

All roles
Staff Engineer
Engineering Lead
Senior SWE
SWE II
SWE I
ML Engineer
Data Engineer

All scores
Top performers (≥ 80)
On track (65–79)
Needs attention (< 65)

{{ filteredCount }} developers

{{ col.label }}

{{ r.chkCell }}
{{ r.rank }}

{{ r.initials }}

{{ r.name }}

{{ r.dept }}
{{ r.teamCell }}

{{ r.role }}
{{ r.overallCell }}
{{ r.v1 }}
{{ r.v2 }}
{{ r.v3 }}
{{ r.v4 }}
{{ r.trendCell }}

Drop in a session — get scored
Drag & drop your Claude Code or Codex session logs (.jsonl), or click to browse. Scored in seconds; your file is deleted right after.
Using Cursor? It has no export file to drag — install the capture agent in Settings and your Cursor sessions upload automatically.

{{ u.badge }}
{{ u.name }}
{{ u.info }}

{{ empInitials }}

{{ empName }}

{{ empRole }}

{{ empDept }}
{{ empStatus }}

{{ empOverall }}

CueBench Score

10-day change
{{ empDelta }}

Sessions scored
{{ empSessionsCount }}

{{ empVectorTitle }}
Anthropic AI Fluency Framework ↗
{{ empRadar }}

{{ empHistoryTitle }}
{{ empDelta }} pts

{{ empHistory }}

{{ v.label }}ⓘ{{ v.def }}Part of Anthropic's AI Fluency Framework — anthropic.com/ai-fluency

{{ v.score }}

Strengths

{{ s.label }}
{{ s.note }}

Where to Coach

{{ s.label }}
{{ s.note }}

No session data — insights will appear once sessions are recorded.

Generating insights…

{{ empInsightsErrorMsg }}

Behavioral signals

{{ sig.mark }}
{{ sig.text }}

View session →

Task performance

{{ tp.category }}
{{ tp.avg }}

{{ tp.category }}
{{ tp.avg }}

{{ empTaskPerfMoreLabel }}

Coaching plan

{{ cp.action }}
{{ cp.why }}

Task breakdown
{{ empTaskPerfExpandLabel }}

{{ tp.category }}
{{ tp.avg }}

Recent sessions

{{ empSessHeaderRow }}

{{ s.id }}{{ s.when }}
{{ s.type }}
{{ s.duration }}
{{ s.tokens }}
{{ s.score }}
{{ s.flagCell }}

{{ showAllSessLabel }}

{{ sessInitials }}

{{ sessId }}

{{ sessQualityLabel }}
ⓘ
{{ sessQualityDef }}

{{ sessTaskLabel }}

{{ sessTask }}

{{ sessEmployee }}
{{ sessModel }}
{{ sessCliLabel }}

{{ sessDate }}
{{ sessTokens }} tokens
◎ Walk through with your coach →

{{ sessScore }}

Session Score

specificity {{ sessSpecificity }}
ⓘ
Specificity measures how clearly and concretely you described what you wanted. High specificity means precise, actionable prompts that name the file, the behavior, and the acceptance condition (e.g. "return 404 instead of 500 when the record id is missing"). Low specificity means vague requests that leave the agent to guess (e.g. "fix it", "make it better").

Session story
◎ Walk through →

{{ storySpan }} · active {{ storyActive }} · away {{ storyIdle }}

{{ pr.mark }}{{ pr.label }}
{{ pr.tip }}

Turn a ✗ into this week's challenge →

+{{ ph.t }}

{{ ph.title }}
{{ ph.chips }}

{{ ph.chev }}

{{ ph.summary }}

“{{ ph.ask }}”

Files  {{ ph.files }}

Activity  {{ ph.activity }}

Interaction trace

{{ ev.t }}

{{ ev.glyph }}

⏸

{{ ev.label }}
{{ ev.detail }}

{{ ev.label }}
{{ ev.detail }}

Score breakdown

{{ b.label }}ⓘ{{ b.def }}Part of Anthropic's AI Fluency Framework — anthropic.com/ai-fluency

{{ b.score }}

AI Insights
Per-axis coaching

Strengths

{{ st.axis }}
{{ st.point }}

Areas to improve

{{ im.axis }}
{{ im.point }}

Configuration
Settings

Account

Name
{{ accountName }}

Email
{{ accountEmail }}

API key

{{ copyApiKeyLabel }}

Your API key identifies your personal workspace — the capture agent and any script can post sessions with it. Treat it like a password.

Connect permanently Optional
Skip manual uploads: a small agent watches your local session logs and sends each finished session automatically.

Run the CueBench capture agent with your API key:

# one-time setup — watches Claude Code, Cursor & Codex sessions (+ any file you put in ~/CueBench Drop)
curl -fsSL {{ agentInstallUrl }} | bash -s -- --key {{ orgApiKey }} --api {{ agentApiUrl }} --origin {{ agentOrigin }}

{{ agentStatusLabel }}
{{ connectCopyLabel }}

Uploading by hand works exactly the same — the agent is just the zero-effort version. It sends session logs only; scoring runs on our servers and files are deleted after scoring. To remove it later: bash ~/.cuebench/capture/uninstall.sh (asks for confirmation).

AI Insights
AI-generated task breakdowns, coaching plans & behavioral signals per session

{{ insightsToggleLabel }}

Refresh every

days

{{ insightsRefreshStatus }}
{{ insightsRefreshLabel }}
{{ insightsSaveLabel }}

Insights are generated by CueBench — no API key needed. They refresh on the cadence above; "Refresh now" regenerates immediately (limited to once every 5 minutes).

Delete my data
Permanently removes your workspace: sessions, scores, coaching history, and survey. This cannot be undone.

{{ deleteDataLabel }}
{{ deleteDataStatus }}

Privacy Policy  ·  Terms of Use

Your Coach
Level up, one session at a time

Focus challenge

{{ coachStatusLabel }}

{{ coachTitle }}
{{ coachPhrasing }}

The technique: {{ coachTechName }}
{{ coachTechHow }}

“{{ coachTechEx }}”

{{ coachTechWhy }}

{{ d.mark }}{{ d.tip }}

{{ coachProgressText }}

{{ coachResolvedText }}
{{ coachNewLabel }}

Your coach picks one measurable thing to improve — based on your weakest skill — and tracks it automatically across your next sessions.

{{ coachNewLabel }}
{{ coachErr }}

Walk through a session
Step-by-step coaching on what happened and what to try instead.

{{ cs.score }}

{{ cs.title }}
{{ cs.when }}

Walk through →

No sessions yet — drop one on your dashboard to get started.

Ask your coach
Answers grounded in your real sessions — scores, traces, habits.

Try: “Why was my discernment low this week?” · “What should I do differently when the agent loops?” · “Which task types am I best at?”

{{ m.content }}

Coach is thinking…

Send

◉
No teams configured
Go to Settings → Teams to create your first team.

Teams

{{ tc.name }}
{{ tc.avgScore }}

{{ tc.memberCount }}

{{ tv.label }}

{{ tv.score }}

← Teams

{{ teamDetailName }}
{{ teamDetailMemberCount }} members

{{ teamDetailScore }}
Team Avg

Team Strengths

{{ ts.label }}
{{ ts.score }}

Where to Coach

{{ tv.label }}
{{ tv.score }}

Behavioral Signals
{{ teamInsightsRefreshLabel }}

Generating team insights…

{{ teamInsightsErrorMsg }}

{{ sig.mark }}
{{ sig.text }}

No AI insights yet. Generate now

Task Performance

No session data yet.

{{ tp.category }}

{{ tp.score }}

{{ tp.category }}

{{ tp.score }}

{{ teamTaskPerfMoreLabel }}

Team Coaching Plan

{{ cp.action }}
{{ cp.why }}

Members

{{ tm.initials }}

{{ tm.name }}
{{ tm.role }}

{{ tm.overall }}

AI-generated · Jun 22, 2026
Reports & Insights

{{ rt.label }}

Executive summary
{{ reportSummary }}

Team score distribution
{{ reportChart }}

30-day vector trends — org average
{{ trendChart }}

{{ rs.value }}
{{ rs.label }}

{{ reportMainTitle }}

{{ ri.title }}
{{ ri.value }}

{{ ri.note }}

Recommendations

{{ rr.urgency }}

{{ rr.action }}
{{ rr.rationale }}