fancysauce · Session naming

See the work behind your AI spend

fancysauce names every Claude Code and Codex session for you. Your usage report shows the work that each dollar paid for, not a list of session IDs.

Claude Code and Codex One plugin install Product preview
Without session naming
7f3c9a1eClaude Code
Mon 3:42 PM · 25m · 12 prompts
≈ $2.10
019a4c2bCodex
Mon 1:15 PM · 41m · 9 prompts
≈ $0.74
With session naming
Fix flaky retry test in billing workerClaude Code
Mon 3:42 PM · 25m · 12 prompts
≈ $2.10
Add CSV export to the invoices pageCodex
Mon 1:15 PM · 41m · 9 prompts
≈ $0.74
The problem

You can see the cost. You cannot see the work.

AI coding tools report sessions by ID. A usage report tells you how many tokens a session used and what it cost. It does not tell you what the developer was doing.

≈ $7.13 across five sessions. Which one fixed the billing bug?
Step 1 · Setup

Install the fancysauce plugin. That is all.

Session naming is part of the fancysauce plugin for Claude Code and Codex. A developer can install it in a minute.

Or IT deploys it to every machine in your organization, with no action from developers.

Claude Code
> /plugin marketplace add FancysauceAI/fancysauce-savings
> /plugin install fancysauce-savings@fancysauce
> /reload-plugins
Codex
$ codex plugin marketplace add FancysauceAI/fancysauce-savings
Managed install · your whole organization
Claude CodeManaged hooks cover every Claude Code session on the machine, with no plugin install.
CodexManaged hooks run for every Codex session on the machine.
Your MDMStep-by-step guides for Jamf Pro, Kandji, and Microsoft Intune.
Step 2 · Most sessions

Most sessions name themselves, at no extra cost

Claude Code gives its own sessions a title. fancysauce picks up that title and shows it in your reports.

No extra model call. No extra cost.

claude — ~/code/acme-app
> the retry test in billing-worker fails about 1 in 5 runs. find out why and fix it
● Read(apps/billing-worker/test/retry.test.ts)
● Bash(npm test -- retry --repeat 20)
⎿ 4 failed, 16 passed
● The test waits 500 ms, but the retry delay can reach 640 ms. I will use fake timers.
● Edit(apps/billing-worker/test/retry.test.ts)
⎿ 20 passed
Title from Claude Code
Fix flaky retry test in billing worker
In fancysauce
Fix flaky retry test in billing workerClaude Code
0 extra model calls$0 naming cost
Result

Every session has a name

Claude Code and Codex wrote four of these titles, at no extra cost. fancysauce wrote the fifth. The next step shows how.

Before
After
Step 3 · The rest

No name? fancysauce writes one when the work is worth it

fancysauce waits until a session has done real work. By default, that is 20,000 tokens. You can tune the threshold for your organization.

A quick question to the assistant stays unnamed and costs nothing extra.

codex — ~/code/acme-app
> add a CSV export button to the invoices page, with a test
• Explored apps/web/src/invoices
• Edited apps/web/src/invoices/InvoiceTable.tsx (+42 −3)
• Ran npm test -- invoices
└ 18 passed
Session work
12,40020,650 tokensthreshold 20,000 · tunable
Title written by fancysauce
Add CSV export to the invoices page

A small, low-cost model writes the title in the background. The developer does not wait.

Rollups

Names roll up into summaries, level by level

Each level summarizes the level below it. Each person's session names become a summary of their week. The people in a department become one department summary. The departments become one summary of the organization.

A manager starts at the top and opens a level only to see more detail.

Organization

31% of AI spend went to the FP&A Q4 forecast refresh. Accounting spent most on the September close.

Department · FP&A

The Q4 forecast refresh, headcount cost variance, and 2027 cash runway scenarios.

Elena M.

The Q4 forecast refresh, and the revenue bridge for the board deck.

Refresh the Q4 revenue forecast with September actualsBuild the revenue bridge chart for the board deck
Jordan T.

Headcount cost variance: plan against actuals, by department.

Compare headcount costs to the planPull September payroll by department
Aisha B.

Cash runway scenarios for the 2027 plan.

Model cash runway under three hiring plansDraft the 2027 plan assumptions memo
Session names
For managers

See what every department is working on

Rollups add one line to the Departments page: what each department is working on.

FP&A spends the most this week. Open it to see why.

Departments

Cost and usage attribution by department · this week
What your organization worked on

31% of AI spend went to the FP&A Q4 forecast refresh. Accounting spent most on the September close. Revenue Operations cleaned up the renewal pipeline.

Total cost$4,821
Month-over-month+12%
Session time700 h
Attribution94%
Department
Working on
Total cost
Composition
FP&A10 people
The Q4 forecast refresh, headcount cost variance, and 2027 cash runway scenarios
$1,482
Accounting9 people
September month-end close: accrual schedules and bank reconciliations
$1,134
Revenue Ops8 people
Renewal pipeline cleanup, and revenue by pricing tier
$891
Procurement6 people
Vendor contract renewals, and a plan to consolidate SaaS spend
$761
Tax5 people
Q3 sales tax filings, and the R&D credit study
$553
Claude CodeCodex
For managers

Open a department to see who is doing what

Each person gets a one-line summary of their week, built from their session names.

Open a person to see the named sessions behind their summary, their spend, and how they split their time.

Departments › FP&A

10 people · 210 session hours · this week
What FP&A worked on

The Q4 forecast refresh with September actuals, a headcount cost variance analysis, and the revenue bridge for the board deck. Cash runway scenarios for the 2027 plan are in progress.

Member
Working on
% time
Tokens
Model mix
Total cost
Elena M.
The Q4 forecast refresh, and the revenue bridge for the board deck
14%
18.2M
$214
Refresh the Q4 revenue forecast with September actualsClaude Code
6%
312K
$3.85
Check deferred revenue against the billing exportCodex
9%
544K
$6.40
Build the revenue bridge chart for the board deckClaude Code
4%
184K
$2.10
+ 41 more sessions
81%
17.2M
$201.65
Jordan T.
Headcount cost variance: plan against actuals, by department
12%
15.8M
$186
Aisha B.
Cash runway scenarios for the 2027 plan
11%
13.0M
$153
Chris W.
Opex budget against actuals, with notes for department heads
9%
10.3M
$121
+ 6 more people
54%
68.7M
$808
OpusSonnetHaikuGPT-6 SolGPT-6 Luna
Cost

You always see what naming costs

fancysauce tracks each title it writes as a separate session. That cost never mixes with your team's work.

  • Naming runs are hidden from the work views and reported as their own line.
  • Each Claude Code naming attempt has a hard cap of $0.02.
  • When a model has no published price, fancysauce shows unpriced. It never shows $0.

Cost & usage

Whole organization · this week
AI coding spend$4,821
Session naming$2.11
Naming share0.04%
Free names71%
Line
Sessions
Cost
Work sessions
812
$4,821
Names from Claude Code and Codex
431
$0
Session naming runs tracked separately
176
$2.11
Naming runs without a published price
3
unpriced
Summary

Every session named. Every department explained.

1

One install

Install the fancysauce plugin, or deploy it across your organization. Naming works from there.

$0

Free names first

fancysauce uses the titles that Claude Code already writes.

20K

Paid titles only after real work

A small, capped model names the sessions that matter, and you see its cost.

4

Levels of rollup

Session, person, department, organization. Start at the top, and open a level for detail.

Session naming is a product preview.