Integrations

One SDK. Every provider. No prompt logging.

Framewren intercepts at the transport layer — token counts, latencies, model identifiers. Your prompt text and completion content are never captured. Add a provider in under 2 minutes; no infrastructure changes, no proxy layer, no ops overhead.

Start monitoring free View quickstart

LLM providers

Every major model provider, one dashboard

Add a provider in under 2 minutes. Framewren normalizes their different pricing models into a unified cost view.

OpenAI

GPT-4o, GPT-4 Turbo, GPT-3.5 Turbo, o1, o3 — all models tracked. Cost per-request and aggregate monthly. Token-level breakdown.

● Connected

Anthropic

Claude 3.5 Sonnet, Claude 3 Opus, Haiku — input/output token costs tracked separately across all context-length tiers.

● Connected

Gemini

Gemini 1.5 Pro, Flash, Ultra — character-based pricing normalized to per-token equivalent for apples-to-apples comparison.

● Connected

Mistral

Mistral Large, Mistral Small, Mixtral 8x7B — latency and cost tracking with region-specific pricing awareness.

● Connected

Cohere

Command R, Command R+, Embed — full cost attribution and latency percentile tracking via the Framewren Python and Node SDKs.

○ Beta

Together AI

Open-source model hosting — Llama 3, Qwen, DeepSeek. Cost per-inference tracked across all model variants and hosting tiers.

○ Beta

Replicate

On-demand model hosting. Track run-level costs by model and version, including cold-start latency overhead separate from inference latency.

○ Coming soon

Self-hosted

vLLM, Ollama, or any OpenAI-compatible API. Framewren tracks latency and infrastructure cost attribution for locally-hosted models.

● Connected

Orchestration frameworks

Works with the LLM frameworks you already use

Drop the Framewren SDK into your LangChain chain, LlamaIndex query engine, or custom HTTP client with zero prompt-layer changes.

LangChain

Callback handler integration

LlamaIndex

Event handler support

Haystack

Pipeline step wrapper

Vercel AI SDK

Edge-compatible middleware

Custom HTTP

Manual SDK init

Setup

Three steps. Under 10 minutes.

No infrastructure changes. No proxy layer. No prompt routing. Just a lightweight SDK that wraps your existing LLM client calls.

Step 1 — Install

bash 
$ pip install framewren-sdk
# or for Node.js:
$ npm install @framewren/sdk

Step 2 — Initialize

python 
import framewren
from openai import OpenAI

framewren.init(api_key="frw_live_your_key_here")
client = framewren.wrap(OpenAI())

Step 3 — You're live

output 
✓ Framewren connected
✓ Monitoring 1 active project: production
  Provider: OpenAI · First metric in 12s

Start monitoring your LLM APIs today

Free tier includes 2 provider connections and 50K API calls per month. No credit card required.

Start free Read the docs