Feature
RCLM captures sessions from local API traffic, enterprise gateway traffic, coding agents, and browser chats. Start with one method, or combine them into a single searchable history. When session encryption is enabled, detailed session blobs are encrypted while metadata remains available for fast product workflows.
Capture methods
Each capture path fits a different workflow, but they all feed the same RCLM account and the same browsing experience.
Run rclm-proxy on your own machine.
Run the local rclm-proxy between your app and providers like Anthropic, OpenAI, Azure OpenAI, and Gemini. Point your requests at a local base URL and every request and response is captured automatically.
The richest capture path for coding workflows.
Install directly into Claude Code and Gemini CLI so RCLM records the full session: messages, tool calls, results, token usage, and every file the agent created or changed.
Capture web chat sessions with no API setup.
Use the Chrome extension to capture conversations from ChatGPT, Claude.ai, and Gemini in the background while you continue using your normal accounts and workflows.
Governed proxy capture for organization traffic.
Enterprise admins can issue team-scoped gateway keys, restrict providers or models, and route production or team API traffic through a hosted gateway. Calls are logged under the correct organization and team.
Normalized record
The important design choice is not just capturing data. It is capturing it into a consistent shape. That is what lets browser chats, local API traces, enterprise gateway calls, and coding-agent sessions live in one dashboard instead of feeling like separate products.
Richer sources contribute richer data, but every capture method still becomes a session you can search, inspect, and export. The detailed blob and the metadata are intentionally separated: encrypted blobs protect raw content, while unencrypted metadata keeps search, summaries, stats, and filters responsive.
Why capture matters
Most people use AI in fragments: browser tabs, coding agents, local API calls, and enterprise gateway traffic. Capture unifies those disconnected traces into one searchable account instead of separate silos.
Not every capture path can see the same data. Browser sessions capture conversation content. Hooks capture the full agent execution trace. The UI stays consistent while the data gets richer where the source allows it.
Use only the extension if that is where you work today. Add the local proxy when you build with APIs. Add hooks when you want coding-session diffs. Add the enterprise gateway when org-level access control matters.
Workflow
01
RCLM intercepts or receives the session from the local proxy, enterprise gateway, hook, or extension.
02
Different raw sources are shaped into a common session record so they can be browsed together.
03
Metadata is indexed for search, stats, and summaries. Full session blobs are stored separately and encrypted when session encryption is enabled.
04
Search, replay, analyze, export, reuse as context, or govern the session in an enterprise workspace.
Where to start
RCLM does not require a wholesale workflow change. The right first step is usually the one that captures your highest-volume AI usage with the least friction.
Once that is in place, you can layer on the other sources and keep building a more complete record over time.
Install the proxy, hooks, or extension now, then bring the rest online when you need a fuller picture. Enterprise teams can add the hosted gateway when they need org-level control over API traffic.