Feature
Find old solutions fast, reopen full conversations, and inspect the exact steps an agent took. Your AI history stops being disposable once it becomes searchable.
Search layer
Search your ChatGPT, Claude, Gemini, proxy, and coding-agent sessions from one place instead of remembering which tool you used last time.
See the whole session, not just the final answer: prompts, responses, tool calls, results, and file changes when the source supports them.
Sessions are titled and described automatically, with tags and other metadata that make browsing easier as your history grows.
Search by topic, then inspect the model, time, token usage, and cost context around the session you are reopening.
Replay view
Search is only useful if opening a result gives you enough detail to act on it. RCLM keeps the session detail view rich enough to answer, "What actually happened here?"
That matters most for coding and debugging workflows, where the missing detail is usually not the final answer. It is the path the agent took to get there.
Why it matters
The practical win is simple: stop re-solving the same problem because a provider-specific chat history buried the answer.
When a coding session goes sideways, replay shows the sequence of prompts, tool usage, and file changes that led there.
Search history turns AI usage from disposable chat into a reusable body of work you can inspect over time.
Workflow
01
Search by phrase, topic, provider, model, date, or other metadata.
02
Jump into the session detail view and inspect the exact conversation context.
03
Review transcripts, tool activity, and file diffs to understand what happened.
04
Export, reference, refine, or add the session to a dataset or workflow.
MCP recall
The ReclaimLLM MCP server lets tools like Claude, Gemini, Cursor, and Codex ask your ReclaimLLM backend for prior sessions without leaving the current workflow.
It is intentionally conservative: search returns short titles and highlights first. A session is only summarized into the current conversation when you explicitly ask to use it.
rclm-hooks-install --with-mcpRun rclm-hooks-install --with-mcp so capture hooks and local MCP recall use the same account configuration.
In a coding assistant, ask whether similar prior work exists or request history for a specific file or folder.
Review the returned titles and highlights, then decide whether any session is worth opening or summarizing.
search_sessions
Find prior bug fixes, performance work, or similar tasks with optional project and file filters
search_by_filename
Find sessions that touched a file or folder when the question is about file history
get_session
Open a specific session summary and link after you provide a session ID
summarize_session
Turn a selected session into reusable context only when you ask for it
Outcomes
Search and replay is where RCLM stops feeling like a storage product and starts feeling like infrastructure for real work.
Once sessions are easy to recover and inspect, they become reusable knowledge instead of expired chat history.
Capture first, then search and replay every useful session instead of starting from scratch each time.