Context Lens: Devtools for your agent context
Solves real LLM cost mystery with zero-code API interception across Claude, GPT, Gemini.
See what your AI sees. Framework-agnostic LLM context window visualizer.
Runs a local proxy and breaks down exactly what's filling your context window — system prompts, tool defs, conversation history, tool outputs and 'thinking' blocks — so you can stop guessing why a session is expensive. It supports Claude, Codex, Gemini, Pi, Aider and more with privacy modes and a background daemon; practical touches like --no-ui and doctor mode show real product thinking. If you care about security, inspect the proxy/key handling before using it with sensitive credentials.
Developers and prompt engineers who use LLMs in their toolchains (e.g., Codex, Claude, Gemini), ML engineers, and teams troubleshooting model cost/behavior
Solves real LLM cost mystery with zero-code API interception across Claude, GPT, Gemini.
SLM classifiers compress context based on tool call intent before LLM sees it.
Subdomain routing means zero /etc/hosts edits—just change base_url and capture everything.
Agents manage their own context window through a transparent local proxy.
Smart context densification for agents, but only macOS and requires agent CLI already installed.
Wire-protocol circuit breaker for agents when LangSmith costs too much.