The temporal codec for agentic perception

Just the deltas,
nothing more.

Every time your AI agent looks at a page, it processes the whole thing. Again. JD Codec is a stateful compression layer that sends only what changed—so your agent sees less, understands more, and acts faster.

0%
Token reduction
1,689k → 230k input tokens
0pp
Success rate gain
66.7% → 83.3%
0%
Faster wall-clock
207.8s → 198.2s across 6 tasks

Same task. Same model. Same tools.

One finishes before
the other starts reading.

Task: Filter products by status

Standard Agent Waiting
Read page state
Find filter control
Apply filter
Verify results
0
tokens
$0.00
cost
0.0s
time
With JD Codec Waiting
Read page state
Find filter control
Apply filter
Verify results
0
tokens
$0.00
cost
0.0s
time

Same model · Same task · Same tools

99% fewer tokens. $2.23 saved per task.

Setup

Three commands.
That's it.

01

Install

One command. No dependencies beyond Python.

pip install jdcodec
02

Connect

Drop-in layer between your agent and the browser. No changes to your agent, model, or tools.

jdcodec start
03

Save

86% fewer input tokens. Higher task success rates. Faster execution. Your agent sees only what changed—and performs better because of it.

Benchmarked

Head-to-head with
standard browser automation.

Same model (Claude Sonnet 4). Same tasks. Same environment. The only difference: standard snapshots vs. JD Codec.

Task Baseline tokens JD Codec tokens Reduction Base JDC
Navigate to Products 108.8k 10.2k -91% PASS PASS
Navigate to Customers 30.6k 8.6k -72% PASS PASS
Navigate to Orders 54.3k 9.9k -82% PASS PASS
Filter Products by Enabled 748.9k 5.2k -99% PASS PASS
Change customer group 203.2k 50.8k -75% FAIL FAIL
Toggle product status 543.5k 145.3k -73% FAIL PASS
Total (6 tasks) 1,689k 230k -86.4% 4/6 5/6

Methodology: 6 real-world tasks on production web app. Input tokens via tiktoken. Wall-clock includes all latency.

Stop burning tokens
on pixels that didn't move.

Get early access to JD Codec. Be the first to know when we launch.