Low-latency execution infrastructure for Polymarket

PolyMach

Low-latency, machine-optimized, persistent Polymarket engine for quantitative and trading agents.

Install Benchmarks GitHub X
Agent path Python SDK -> native runtime 
uv add polymach

First use downloads the correct signed optimized runtime for the host machine, verifies it, starts a persistent local session, and gates usage through a machine-bound license.

Live benchmark snapshot

Latency profile

Numbers below are pulled from the public SDK README benchmark run against the official polymarket-cli. Values are p50 latency on March 8, 2026. Lower is better.

PolyMach Official CLI p50 latency ยท lower is better

Quote path

Live book quote

3554.7x faster
PolyMach
0.022 ms
Official CLI
78.48 ms

Midpoint path

Midpoint lookup

3801.1x faster
PolyMach
0.020 ms
Official CLI
76.70 ms

Discovery path

Market discovery

4.7x faster
PolyMach
64.12 ms
Official CLI
300.46 ms

Account path

Account health equivalent

5.0x faster
PolyMach
54.84 ms
Official CLI
271.84 ms

Execution path

Order create

6.6x faster
PolyMach
28.60 ms
Official CLI
188.37 ms

Execution path

Order cancel

4.4x faster
PolyMach
28.43 ms
Official CLI
124.31 ms

Stream path

Stream cadence vs polling

1.5x faster
PolyMach
51.15 ms
Official CLI
75.65 ms

Execution model

Built around a hot local session

The public SDK talks to a warmed local Rust runtime with persistent state, prewarmed token context, websocket-backed data, and low-overhead repeat calls. The runtime is tuned around hot-path micro-optimizations in quote access, request construction, cache reuse, and repeat-call overhead.

Capabilities

  • Discover markets and resolve token ids
  • Stream quotes, midpoint, and live market state
  • Read balances, positions, and account health
  • Post and cancel limit orders through the persistent engine
Signed runtime manifest and artifacts
Native target-specific runtime selection
Persistent local session for low-latency repeat calls

Minimal operating model

What actually happens

01

Install the SDK

uv add polymach installs the public Python package only.

02

Bootstrap the runtime

The SDK detects the machine, fetches the matching signed native runtime, verifies it, and caches it locally.

03

Pay, activate, trade

The machine activates against the license flow, then the client uses the persistent engine for discovery, quotes, streams, and order execution.

Design intent

Designed around the runtime

Public Python surface. Private optimized Rust runtime. Signed release catalog. Persistent local process. Fast repeat-path execution. Native target-specific builds and low-level micro-optimizations across warmed data paths, session reuse, and execution plumbing. Built for agents that need persistent market execution infrastructure.

Payment

50 USDC per machine cycle

PolyMach uses a flat subscription model: 50 USDC per machine for each 30-day cycle. Payment is made on Polygon in USDC, and the SDK can either return exact payment instructions or submit the payment directly from a configured signer.

Agents can prepay multiple cycles in one transaction and activate the local machine against that payment.

More information

Read the SDK

For the public Python SDK, installation details, API surface, and documentation, see the GitHub repository.

github.com/polymach-dev/polymach

Follow

Follow Us on X

For updates, release notes, and product development, follow PolyMach on X.

x.com/PolyMachQuant