Lynkr
Make claude Code work with any LLM
FreeCLIOpen SourceCode572 impressions#1371 of its week
More CLI tools →74 free dev tools →
Lynkr is a self-hosted proxy that compresses and routes AI coding tool requests to any LLM provider.
- for
- Developers using Claude Code or similar AI coding tools who want lower token costs and multi-model support.
- pricing
- open source
- license
- Apache-2.0
Key features
- Repo intelligence — Builds a lightweight index of workspace files, symbols and references to improve model context.
- Tool call support — Provides sandboxed tasks, file edits, script execution and custom tool manifests.
- Git-integrated workflows — AI-assisted diff review, commit message generation and selective staging with auto-commit helpers.
- Prompt caching — Semantic cache returns near-identical prompts in ~171 ms with zero tokens billed.
- Smart tool selection — Strips irrelevant tool schemas per request, reducing token usage by up to 53 %.
- Complexity tier routing — Scores requests on 15 dimensions and routes simple queries to cheap local models, complex ones to cloud providers.
- MCP Code Mode — Replaces 100+ tool schemas with 4 meta-tools, cutting MCP token overhead dramatically.
- Multi-provider endpoint — One local endpoint forwards to 13+ LLM providers (local, cloud, enterprise).
Use cases
- Reduce token costs for Claude Code by compressing JSON tool results before sending to the model.
- Run AI coding agents on a local LLM (e.g., Ollama) for cheap, fast feedback on simple tasks.
- Automatically route security-sensitive or complex code queries to enterprise LLMs while keeping routine work local.
- Cache repeated prompts in CI pipelines to eliminate token billing for identical queries.
Lynkr vs alternatives
 Lynkr |  Augment Code |  OpenCode |  Cursor |  Continue.dev | |
|---|---|---|---|---|---|
| Best for | Token-efficient LLM gateway | Deep codebase context | Terminal AI coding agent | AI code editor | IDE AI assistant |
| Pricing | Open source | Subscription | Free | Subscription | Free |
| DevHunt upvotes | 0 | 0 | 0 | 0 | 0 |
| Launched | Jan 2026 | — | — | — | — |
- Lynkr vs Augment Code: Augment Code focuses on deep codebase context but does not provide token compression or multi-provider routing.
- Lynkr vs OpenCode: OpenCode is an open-source terminal agent without Lynkr's proxy-level token reduction and tier routing.
- Lynkr vs Cursor: Cursor is an AI-powered editor; Lynkr is a separate gateway that works with any editor.
- Lynkr vs Continue.dev: Continue.dev offers IDE assistance but lacks Lynkr's compression, caching, and multi-LLM endpoint.
Lynkr FAQ
Do I need to modify my existing AI coding tool code to use Lynkr?+
No. Lynkr works via a single environment variable pointing the tool at the Lynkr endpoint; no code changes are required.
Which LLM providers does Lynkr support?+
Lynkr can forward requests to 13+ providers, including local models (Ollama, llama.cpp, LM Studio) and cloud services like Azure OpenAI, OpenAI, Anthropic, Bedrock, OpenRouter, Databricks, and Vertex AI.
How does Lynkr reduce token usage?+
It compresses structured JSON tool results, strips unused tool schemas, and caches near-identical prompts, achieving up to 87.6 % token reduction on JSON data.
Is Lynkr free to use?+
Yes. Lynkr is released under the Apache 2.0 license, self-hosted, and has no usage tracking or mandatory fees.
Can Lynkr handle multiple agents simultaneously?+
Yes. The gateway can spawn multiple agents at once and routes each request based on its complexity tier.
What is the performance impact of the compression and caching layers?+
Optimizations run in-process with negligible latency; cache hits return in ~171 ms, and compression adds no meaningful delay.
Summarized by DevHunt from fast-editor.github.io · Oct 26, 2026. Details may change; check the official site.
