Lynkr

Run Claude Code, Cursor, and Codex on any model. One proxy, every provider.

12+ LLM Providers

60-80% Cost Reduction

699 Tests Passing

0 Code Changes Required

---

The Problem

AI coding tools lock you into one provider. Claude Code requires Anthropic. Codex requires OpenAI. You can't use your company's Databricks endpoint, your local Ollama models, or your AWS Bedrock account — at least, not without Lynkr.

The real costs:

• Anthropic API at $15/MTok output adds up fast for daily coding
• No way to use free local models (Ollama, llama.cpp) with Claude Code
• Enterprise teams can't route through their own cloud infrastructure
• Provider outages take your entire workflow down

The Solution

Lynkr is a self-hosted proxy that sits between your AI coding tools and any LLM provider. One environment variable change, and your tools work with any model.

Claude Code / Cursor / Codex / Cline / Continue / Vercel AI SDK
                        |
                      Lynkr
                        |
    Ollama | Bedrock | Databricks | OpenRouter | Azure | OpenAI | llama.cpp

# That's it. Three lines.
npm install -g lynkr
export ANTHROPIC_BASE_URL=http://localhost:8081
lynkr start

---

Quick Start

Install

npm install -g pino-pretty && npm install -g lynkr

Pick a Provider

Free & Local (Ollama)

export MODEL_PROVIDER=ollama
export OLLAMA_MODEL=qwen2.5-coder:latest
lynkr start

AWS Bedrock (100+ models)

export MODEL_PROVIDER=bedrock
export AWS_BEDROCK_API_KEY=your-key
export AWS_BEDROCK_MODEL_ID=anthropic.claude-3-5-sonnet-20241022-v2:0
lynkr start

OpenRouter (cheapest cloud)

export MODEL_PROVIDER=openrouter
export OPENROUTER_API_KEY=sk-or-v1-your-key
lynkr start

Connect Your Tool

Claude Code

export ANTHROPIC_BASE_URL=http://localhost:8081
export ANTHROPIC_API_KEY=dummy
claude "Your prompt here"

Codex CLI — edit ~/.codex/config.toml:

model_provider = "lynkr"
model = "gpt-4o"

[model_providers.lynkr]
name = "Lynkr Proxy"
base_url = "http://localhost:8081/v1"
wire_api = "responses"

Cursor IDE

• Settings > Features > Models
• Base URL: http://localhost:8081/v1
• API Key: sk-lynkr

Vercel AI SDK

import { generateText } from "ai";
import { createOpenAICompatible } from "@ai-sdk/openai-compatible";

const lynkr = createOpenAICompatible({
  baseURL: "http://localhost:8081/v1",
  name: "lynkr",
  apiKey: "sk-lynkr",
});

const { text } = await generateText({
  model: lynkr.chatModel("auto"),
  prompt: "Hello!",
});

Works with any OpenAI-compatible client: Cline, Continue.dev, ClawdBot, KiloCode, and more.

---

Supported Providers

Provider	Type	Models	Cost
Ollama	Local	Unlimited (free, offline)	Free
llama.cpp	Local	Any GGUF model	Free
LM Studio	Local	Local models with GUI	Free
MLX Server	Local	Apple Silicon optimized	Free
AWS Bedrock	Cloud	100+ (Claude, Llama, Mistral, Titan)	$$
OpenRouter	Cloud	100+ (GPT, Claude, Llama, Gemini)	$-$$
Databricks	Cloud	Claude Sonnet 4.5, Opus 4.6	$$$
Azure OpenAI	Cloud	GPT-4o, o1, o3	$$$
Azure Anthropic	Cloud	Claude models	$$$
OpenAI	Cloud	GPT-4o, o3, o4-mini	$$$
Google Vertex	Cloud	Gemini 2.5 Pro/Flash	$$$
Moonshot AI	Cloud	Kimi K2 Thinking/Turbo	$$
Z.AI	Cloud	GLM-4.7	$$
DeepSeek	Cloud	DeepSeek Reasoner, R1	$

4 local providers for 100% offline, free usage. 10+ cloud providers for scale.

---

Why Lynkr Over Alternatives

Feature	Lynkr	LiteLLM (42K stars)	OpenRouter	PortKey
Setup	npm install -g lynkr	Python + Docker + Postgres	Account signup	Docker + config
Claude Code support	Drop-in, native	Requires config	No CLI support	Requires config
Cursor support	Drop-in, native	Partial	Via API key	Partial
Codex CLI support	Drop-in, native	No	No	No
Built for coding tools	Yes (purpose-built)	No (general gateway)	No (general API)	No (general gateway)
Local models	Ollama, llama.cpp, LM Studio, MLX	Ollama only	No	No
Token optimization	Built-in (60-80% savings)	No	No	Caching only
Complexity routing	Auto-routes by task difficulty	Manual	Cost/latency only	Manual
Memory system	Titans-inspired long-term memory	No	No	No
Self-hosted	Yes (Node.js)	Yes (Python stack)	No (SaaS)	Yes (Docker)
Offline capable	Yes	Yes	No	No
Transaction fees	None	None (OSS) / Paid enterprise	5.5% on credits	Free tier / Paid
Dependencies	Node.js only	Python, Prisma, PostgreSQL	N/A	Docker, Python
Format conversion	Anthropic <-> OpenAI (automatic)	Automatic	N/A	Automatic
Code intelligence	Graphify (19-lang AST graph)	No	No	No
Routing telemetry	Built-in (SQLite + REST API)	No	Dashboard	Dashboard
Admin hot-reload	Yes (no restart)	Requires restart	N/A	Requires restart
License	Apache 2.0	MIT	Proprietary	MIT (gateway)

Lynkr's edge: Purpose-built for AI coding tools. Not a general LLM gateway — a proxy that understands Claude Code, Cursor, and Codex natively, with built-in token optimization, complexity-based routing, and a memory system designed for coding workflows. Installs in one command, runs on Node.js, zero infrastructure required.

---

Cost Comparison

Scenario	Direct Anthropic	Lynkr + Ollama	Lynkr + OpenRouter	Lynkr + Bedrock
Daily Claude Code usage	~$10-30/day	$0 (free)	~$2-8/day	~$5-15/day
Token optimization savings	—	—	60-80% further	60-80% further
Monthly (heavy use)	$300-900	$0	$60-240	$150-450

With token optimization enabled, Lynkr's smart tool selection, prompt caching, and memory deduplication reduce token usage by 60-80% on top of provider savings.

---

What's Under the Hood

Lynkr isn't just a passthrough proxy. It's an optimization layer.

Smart Routing (5-Phase)

Routes requests to the right model based on 5-phase complexity analysis. Simple questions go to fast/cheap models. Complex architectural tasks go to powerful models. Includes Graphify structural analysis for code-aware routing.

• Complexity scoring — 15-dimension weighted scoring with agentic workflow detection
• Graphify integration — AST-based knowledge graph detects god nodes, community cohesion, blast radius across 19 languages
• Routing telemetry — every decision recorded with quality scoring (0-100) and latency tracking (P50/P95/P99)

Token Optimization (7 Phases)

• Smart tool selection — only sends tools relevant to the current task
• Code Mode — replaces 100+ MCP tools with 4 meta-tools (~96% token reduction)
• Distill compression — structural similarity, delta rendering, smart dedup of repetitive tool outputs
• Prompt caching — SHA-256 keyed LRU cache
• Memory deduplication — eliminates repeated information across turns
• History compression — sliding window with Distill-powered structural dedup
• Headroom sidecar — optional 47-92% ML-based compression (Smart Crusher, CCR, LLMLingua)

Enterprise Resilience

• Circuit breakers — automatic failover with half-open probe recovery
• Admin hot-reload — POST /v1/admin/reload reloads config + resets circuit breakers without restart
• Load shedding — graceful degradation under high load
• Prometheus metrics — full observability at /metrics
• Health checks — K8s-ready endpoints at /health
• Performance timer — per-request timing breakdown with PERF_TIMER=true

Memory System

Titans-inspired long-term memory with surprise-based filtering. The system remembers important context across sessions and forgets noise — reducing token waste from repeated context.

Semantic Cache

Cache responses for semantically similar prompts. Hit rate depends on your workflow, but repeat questions (common in coding) get instant responses.

SEMANTIC_CACHE_ENABLED=true
SEMANTIC_CACHE_THRESHOLD=0.95

MCP Integration + Code Mode

Automatic Model Context Protocol server discovery and orchestration. Your MCP tools work through Lynkr without configuration. Enable Code Mode to replace 100+ MCP tool definitions with 4 lightweight meta-tools:

CODE_MODE_ENABLED=true  # ~96% reduction in tool-catalog tokens

---

Deployment Options

NPM (recommended)

npm install -g lynkr && lynkr start

Docker

docker-compose up -d

Git Clone

git clone https://github.com/Fast-Editor/Lynkr.git
cd Lynkr && npm install && cp .env.example .env
npm start

Homebrew

brew tap vishalveerareddy123/lynkr
brew install lynkr

---

Documentation

Guide	Description
Installation	All installation methods
Provider Config	Setup for all 12+ providers
Claude Code CLI	Detailed Claude Code integration
Codex CLI	Codex config.toml setup
Cursor IDE	Cursor integration + troubleshooting
Embeddings	@Codebase semantic search (4 options)
Token Optimization	60-80% cost reduction strategies
Memory System	Titans-inspired long-term memory
Tools & Execution	Tool calling and execution modes
Smart Routing	Complexity-based model routing
Docker Deployment	docker-compose with GPU support
Production Hardening	Circuit breakers, metrics, load shedding
API Reference	All endpoints and formats
Troubleshooting	Common issues and solutions
FAQ	Frequently asked questions

---

Troubleshooting

Issue	Solution
Same response for all queries	Disable semantic cache: SEMANTIC_CACHE_ENABLED=false
Tool calls not executing	Increase threshold: POLICY_TOOL_LOOP_THRESHOLD=15
Slow first request	Keep Ollama loaded: OLLAMA_KEEP_ALIVE=24h
Connection refused	Ensure Lynkr is running: lynkr start

---

Contributing

We welcome contributions. See the Contributing Guide and Testing Guide.

---

License

Apache 2.0 — See LICENSE.

---

Community

• GitHub Discussions — Questions and tips
• Report Issues — Bug reports and feature requests
• NPM Package — Official package
• DeepWiki — AI-powered docs search

---

Built by Vishal Veera Reddy — for developers who want control over their AI tools.