teamchong/pxpipePublic

Cut Claude Code's input tokens by rendering bulky context as images — the same system prompt, tool docs, and history, in a fraction of the tokens.

An image's token cost is fixed by its pixel dimensions, not by how much text is inside it. Dense content (code, JSON, tool output) packs ~3.1 chars per image-token vs ~1 char per text-token on real Claude Code traffic. The reader is the same vision channel that Anthropic's computer use already relies on for screenshots. pxpipe is a local proxy that uses that channel for context: it rewrites the bulky parts of each request into compact PNGs before it leaves your machine. At current Fable list prices that lands as a ~59–70% lower end-to-end bill — but prices move and workloads differ, so the durable number is the token cut itself, measured per-request against a free count_tokens counterfactual in ~/.pxpipe/events.jsonl.

This is what the model sees instead of text:

~48k chars of system prompt + tool docs: ≈25k tokens as text, ≈2.7k image tokens as this page. Real pipeline output; the model reads renders like this at 100/100 (see benchmarks).

Eight years of context growth, in characters. Every text line tops out near ~4M chars (a 1M-token window at ~4 chars/token); Grok 4.5 is shown as a text-window point only (500K). The orange overlays are the same 1M windows read through pxpipe images — ~19.0M chars for Fable 5 (4.8×) and ~21.3M chars for Gemini 3.6 Flash (5.3× text capacity). Density is measured from a live render at generation time, not hand-typed: regenerate with npx tsx scripts/gen-context-chart.ts (source).

Fable 5 (the default, 100/100 reader) — plain left, pxpipe right:

pxpipe counts an exact token 10/10 across 39 imaged filler files (matches grep line-for-line), gets the multi-step ledger arithmetic right, and ends the session at $6.06 with context to spare (73.5k/1M) vs $42.21 at 96% full. One caveat visible in the clip: the pxpipe arm needed a nudge to match the requested one-line output format.

npx pxpipe-proxy                                  # proxy on 127.0.0.1:47821
ANTHROPIC_BASE_URL=http://127.0.0.1:47821 claude  # point Claude Code at it

Dashboard at http://127.0.0.1:47821/: tokens saved, every text→image conversion side by side, kill switch, live model chips. Responses stream normally — pxpipe compresses the request only, never the model's output. Recent turns stay text; the system prompt, tool docs, and older bulk history are imaged.

pxpipe warp -- claude          # also: cursor-agent, codex, or a shell alias

Same thing without ANTHROPIC_BASE_URL, so /remote-control, claude.ai connectors, and first-party gates keep working. Full instructions in the dashboard.

api.anthropic.com/v1/messages is routed by default. Agents that reach their provider over some other base URL need a rule for it, and a rule that names a port matches only that port:

pxpipe warp --route '127.0.0.1:9090/v1/*=http://127.0.0.1:47821' -- codex

You can render text, files, or diffs to PNG pages without running the proxy or connecting Claude Code:

npx pxpipe-proxy export src/
cat prompt.txt | npx pxpipe-proxy export --stdin
npx pxpipe-proxy export --git

If the package is installed, use pxpipe export instead of npx pxpipe-proxy export.

Each run writes a fresh pxpipe-export-XXXXXX/ output folder (the exact path is printed when the command finishes) containing page-*.png, factsheet.txt, manifest.json, and prompt.txt. Upload the PNG pages and paste the prompt into image-upload clients such as Cursor when you want dense visual context without running the proxy.

• It is lossy. Exact 12-char hex strings in dense imaged content: 13/15 on Fable 5 and 0/15 on Sol — misses are silent confabulations, not errors. Byte-exact values (IDs, hashes, secrets) must stay text; recent turns do. The factsheet selectively preserves up to 96 recognized precision-critical tokens, not every identifier. A dedicated verbatim-risk guard is not built yet.
• Escape hatch: subagents on non-allowlisted models pass through as text — route byte-exact work there (CLAUDE_CODE_SUBAGENT_MODEL=claude-sonnet-4-6, or model: sonnet in agent frontmatter).
• Real work: SWE-bench Lite pilot 10/10 both arms at −65% request size; SWE-bench Pro 14/19 ON vs 15/19 OFF at −60%, verdicts agree 18/19, and the single split re-resolved 3/3 on replication — run-to-run variance, not compression. Small n; receipts in eval/.
• Workload-dependent. Wins on token-dense content (~1 char/token), loses money on sparse prose (~3.5 chars/token); a profitability gate (calibrated on N=391 production rows) images only where the math wins.
• Client-dependent. Savings track uncached bulk the client still re-sends as text. Claude Code re-sends system + tools + history on /anthropic/messages and typically lands ~60–70%. Details and measured splits: docs/CACHING_AND_SAVINGS.md.

This matrix shows coverage as well as scores. — means the model was not run on that test; it does not mean zero. Arithmetic uses novel random-number problems. Gist, state, and never-stated probes share one corpus. Never-stated is confabulations, so lower is better. The numbers at column is the render geometry the row's scores were measured at; a model's shipped profile can differ (Sol and Qwen ship the measured 14px/84 geometry, but their broad-suite numbers predate it).

Offline export of the same deterministic 454,045-character dense record corpus through each complete profile produced:

The text estimate uses 3.7 characters/token; image tokens use each model's provider formula and actual rendered page dimensions. These figures establish profile cost on this corpus, not a universal workload savings rate. Sol's paid fixtures estimated 42% while reading 7/8 exact with no unsupported inventions.

The runs use different transports and profile generations, not one identical image geometry. Fable and Opus use Claude; Gemini uses Google AI Studio; Sol and Grok use Codex Responses; Kimi K3 uses Cloudflare's OpenAI-compatible transport. Current production profiles include the adjacent bounded factsheet; historical or pure-image exceptions are identified in the linked evaluation.

These are not cross-model comparisons. Every unlisted model is not run.

The SWE-bench runner is Claude Code/Fable-specific; no other model has an ON/OFF run. Gemini's positional-retrieval sweep is directional evidence, not a general Lost-in-the-Middle result.

Measured by rendering this repo’s dense fixture through the real pipeline and pricing pixels at each family’s vision rate. Multiplier = measured chars/vision-token ÷ 4 (prose text baseline). Not a model-quality score.

Regenerate: npx tsx scripts/gen-context-chart.ts · chart PNG docs/assets/context-window-chars.png.

The older GSM8K result is omitted because its training-data contamination can hide image misreads; the linked arithmetic evaluations use novel numbers.

model id ──► render profile ──► wrap/reflow bulk context ──► PNG[] + bounded factsheet

The proxy handles Anthropic Messages, OpenAI Responses and Chat Completions, and Google generateContent requests. It rewrites eligible bulk into image blocks and forwards the provider-native request, or bridges Anthropic Messages to a configured OpenAI-compatible provider. On Anthropic, the static prefix and prompt-cache boundary are preserved. Model-specific profiles control geometry, factsheets, history retention, and profitability, so sparse prose stays text. Events log to ~/.pxpipe/events.jsonl.

import { renderTextToImages, transformAnthropicMessages } from "pxpipe-proxy";

const { pages } = await renderTextToImages(toolResultText);     // pages[i].png: Uint8Array
const { body, applied, info } = await transformAnthropicMessages({
  body: requestBytes,
  model: "claude-fable-5",
});

options.keepSharp(block) pins blocks as text; options.emitRecoverable returns the originals of imaged blocks. Pure-JS runtime (Node and edge/Workers); @napi-rs/canvas is build-time only. Full API: src/core/index.ts.

pxpipe stats                   # human report from ~/.pxpipe/events.jsonl
pxpipe stats --json            # same aggregate as machine-readable JSON
pxpipe stats --file /path/to/events.jsonl

pnpm install && pnpm test
pnpm run build                # regenerates dist/

Windows is community-supported: primary development targets macOS/Linux, and Windows-specific fixes rely on contributor PRs (thanks @makoribrian).

• PNG encoding adds latency to large requests before they leave.
• ASCII/Latin-1 well tested; CJK works but conservatively.

Current as of 2026-07-22. The broad conclusion from the 2026-07-05 pass still holds: exact recall is limited by pixels per glyph, so rendering changes do not eliminate errors at profitable density. A later glyph-style A/B did find a useful local improvement: repainting K reduced Fable's H/K error from 47.2% to 18.7% without changing geometry or token cost. It shipped, but exact control IDs did not improve. See FINDINGS.md, 2026-07-19 entry.

Runtime canary + text re-fetch and surrogate-reader pre-flight remain untested. The release tripwire remains a resolution sweep for each new model; a model that reads production cells near 100% would permit higher density.

Effective-context benefits remain unproven. The production-history results above are directional evidence, not a general context-window or long-task accuracy claim.

Third-party projects listed here are not maintained or supported by pxpipe.

• pxpipe-windows — Windows support for pxpipe mitm (node-forge CA in place of openssl, Task Scheduler autostart).
• OmniGlyph — A community-maintained project derived from pxpipe and used by OmniRoute.
• pxpipe-go — A Go port of pxpipe's core with a CLI wrapper, standalone proxy, and embeddable library for Anthropic Messages and OpenAI Chat/Responses.

MIT.