prune.

Documentation

Technical guides for the Prune LLM proxy — integration, security, and comparisons.

Quick answer

Prune is a managed AI gateway. Change your SDK base URL, add a Prune API key, and every response includes a cost/security receipt. Unique-path savings apply even when prompts never repeat; cache is upside. Try the free Prune Check tool below, or see how Prune compares to LiteLLM, cache-first gateways, Portkey, Helicone, and Cloudflare AI Gateway.

Prune Check

Free tool: paste a prompt, see security flags, cost, and a sample receipt. No signup.

API coding tools

Optional: OpenAI-compatible / OpenRouter clients through Prune.

How it works

Request lifecycle, Shield, unique-path savings, prune_metadata receipts.

Multi-endpoint apps (Focus-style)

X-Prune-Workload, X-Prune-Feature-Id, Recommended vs Max, first-miss expectations.

Security model

AES-256-GCM vault, fail-closed Shield, canary tokens, audit fixes.

Prune vs cache-first gateways

Why hit rate is not the thesis — unique-path + Shield + proof.

Prune vs LiteLLM

Self-hosted vs managed — when to use each.

Prune vs Portkey

Guardrails-as-plugin vs default-on Shield enforcement.

Prune vs Helicone

Observability-first vs a control plane that enforces policy and cost.

Prune vs Cloudflare AI Gateway

Edge caching/analytics vs a provider-agnostic security + cost layer.

Prune is in private beta — request early access on the homepage. Full dashboard access unlocks after your invite.

Quick start

AES-256-GCM encryption for vaulted provider API keys vault · ≥25% median cost reduction on unique (non-cache) traffic in the published benchmark suite · Spend caps enforced in under 10ms, before the provider call

quickstart.py
pip install prune-sdk

from prune import OpenAI

client = OpenAI(
    api_key="sk-...",
    prune_api_key="prune_...",  # from dashboard → API Keys
)

resp = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(resp.choices[0].message.content)
print(client.last_prune_metadata)  # cache_hit, tokens_saved, ...