Raft agent usage

2026-07-26 to 2026-08-12 EDT · generated 2026-08-12 11:40

$1,148

notional API-equivalent across 11 agents over 18 days

These agents run on a Claude subscription, not a metered API key. Every dollar here is a modelled API-equivalent at published list rates, useful for comparing agents against each other. It is not an invoice and nobody billed it.

$63.76
Per day
fleet average over the window
1,329
Jobs dispatched
$0.86 per job
2.0B
Tokens billed
97.0% of them cache reads
$188
Heaviest day
2026-08-09

Where the money goes

97% of every token billed is the fleet re-reading its own context

Cache reads are the cheapest token there is, at a tenth of the input rate, and they still account for $693 of the total. Cache writes add $395, and everything the agents actually wrote comes to $59.66. Shrinking what an agent carries between turns moves this number; talking less does not.

  • Cache reads$69360.4%
  • Cache writes (1h)$39534.4%
  • Output$59.665.2%
  • Fresh input$0.040.0%

Daily

Fleet cost per day

Notional cost across all agents, by calendar day.

$0.00$100$200$30007-2607-2807-3008-0108-0308-0508-0708-0908-1108-12

Per agent, per day

Every agent keeps a different rhythm

Panels share one y-scale, so height means the same thing in all of them and the quiet agents read as quiet. Switch to the per-agent scale to see the shape of a small agent's week without the loudest one flattening it.

SEO-Ops$487

peak $98.54 on 08-02 · 13/18 days active

Mika$189

peak $43.51 on 08-08 · 12/18 days active

Cindy$184

peak $35.44 on 08-11 · 18/18 days active

Outbound-Ops$138

peak $31.28 on 08-11 · 12/18 days active

PPC-Ops$112

peak $33.77 on 08-09 · 13/18 days active

Dev-Ops$14.46

peak $12.37 on 08-08 · 3/18 days active

Client-Reporting$9.37

peak $5.25 on 08-09 · 4/18 days active

Rex$6.31

peak $4.71 on 08-08 · 3/18 days active

Nova$4.46

peak $2.47 on 08-08 · 4/18 days active

Aria$3.24

peak $1.92 on 08-08 · 3/18 days active

Spy-Ops$1.12

peak $1.12 on 08-11 · 1/18 days active

By agent

SEO-Ops alone is 42% of the fleet at $487

Ranked by notional cost. The bar is the row, so length and number carry the same fact.

  • SEO-Ops$487
    42.4% of fleet · 152 jobs · 1,302 messages · 660M tokens · Opus 4.8 · Opus 5
  • Mika$189
    16.4% of fleet · 184 jobs · 626 messages · 214M tokens · Opus 4.8
  • Cindy$184
    16.0% of fleet · 405 jobs · 1,141 messages · 509M tokens · Sonnet 5 · Haiku 4.5
  • Outbound-Ops$138
    12.0% of fleet · 315 jobs · 1,089 messages · 413M tokens · Sonnet 5 · Opus 4.8
  • PPC-Ops$112
    9.7% of fleet · 72 jobs · 408 messages · 119M tokens · Opus 5 · Opus 4.8
  • Dev-Ops$14.46
    1.3% of fleet · 39 jobs · 159 messages · 20.3M tokens · Opus 4.8 · Haiku 4.5
  • Client-Reporting$9.37
    0.8% of fleet · 71 jobs · 193 messages · 26.3M tokens · Sonnet 5 · Opus 4.8
  • Rex$6.31
    0.6% of fleet · 34 jobs · 80 messages · 7.2M tokens · Opus 4.8 · Opus 5
  • Nova$4.46
    0.4% of fleet · 37 jobs · 96 messages · 10.1M tokens · Sonnet 5 · Opus 4.8
  • Aria$3.24
    0.3% of fleet · 11 jobs · 34 messages · 3.2M tokens · Opus 4.8 · Sonnet 5
  • Spy-Ops$1.12
    0.1% of fleet · 9 jobs · 37 messages · 2.1M tokens · Opus 5 · Haiku 4.5

Detail

Every agent, every bucket

AgentJobsMessagesCache readsCache writesOutputCostPer jobActive
SEO-Ops1521,302644M15.4M1.1M$487$3.202026-07-26 → 2026-08-11
Mika184626206M7.4M477K$189$1.022026-07-26 → 2026-08-08
Cindy4051,141488M21.3M272K$184$0.452026-07-26 → 2026-08-12
Outbound-Ops3151,089408M4.6M373K$138$0.442026-07-26 → 2026-08-12
PPC-Ops72408113M5.2M337K$112$1.552026-07-26 → 2026-08-11
Dev-Ops3915919.8M392K89.1K$14.46$0.372026-08-02 → 2026-08-11
Client-Reporting7119325.5M773K48.2K$9.37$0.132026-08-02 → 2026-08-11
Rex34806.9M278K23.9K$6.31$0.192026-07-26 → 2026-08-09
Nova37969.7M394K18.2K$4.46$0.122026-07-26 → 2026-08-11
Aria11343.0M256K12.3K$3.24$0.292026-07-26 → 2026-08-09
Spy-Ops9371.9M166K9.1K$1.12$0.122026-08-11 → 2026-08-11

Models

What each model cost the fleet

Synthetic records are Claude Code's own local messages. They carry no model and no charge, so they are counted and priced at zero.

ModelMessagesTokensCostShare
Opus 4.82,105714M$59651.9%
Sonnet 52,047914M$29125.4%
Opus 5826346M$25822.5%
Haiku 4.510510.3M$2.440.2%
Synthetic (unbilled)820$0.000.0%

Method

How these numbers were built

Source

Claude Code transcripts under ~/.claude/projects/*slock-agents-*, one scope per agent. Names come from the first line of each agent's MEMORY.md.

Deduplication

A usage record repeats once per content block of the same assistant message. Counting raw lines overstates tokens by roughly 80%, so every message is counted once, keyed on its message id.

Cache pricing

Writes are split by TTL because the price differs: a 1-hour write costs 2x the input rate against 1.25x for 5 minutes. This fleet writes 1-hour caches almost exclusively.

Rates

Published list rates read from the pricing docs on 2026-08-11, held in rates.json. Fast mode, US-pinned inference and batch tier are all applied when a record carries them.