txcript: an MCP server for portable AI conversation sessions
txcript, developed by Skillsynchq, is a Model Context Protocol server and session conversion engine that moves full AI conversations between agent platforms. The tool exports and imports entire sessions while preserving messages, tool calls, and reasoning so users can continue work in a different assistant. Key capabilities include cross-harness portability, local searchable session memory, and transcript mapping across agent formats. It targets developers, AI power users, and researchers who need persistent, machine-accessible conversation history.
What tasks can you actually use it for?
Use the tool to migrate active chat sessions between different AI assistants and IDE-integrated agents, keeping the conversation state intact. It maps messages, tool calls, tool results, and the agent's internal reasoning to a typed common model, and supports more than 16 harness formats including Claude Code, Cursor, Grok, and Codex. For users juggling multiple agents, that mapping lets a session continue in another harness without losing hidden state.
How accurate are conversions and search?
Conversions are designed to be precise: byte-lossless conversion reproduces the original file exactly when saved back in native format. Session search runs locally and uses a high-performance interactive index powered by nucleo, so queries scan the machine's stored sessions rather than relying on a remote index. Those two elements address both fidelity of export/import and fast local retrieval of past interactions.
What inputs and environments does it accept?
The tool operates where the Model Context Protocol is available, making it compatible with MCP-supporting harnesses across common desktop environments. It recognizes the transcript shapes used by many agent formats and converts between them using its central interchange model. Supported harness examples span code assistants and agent platforms that developers commonly embed in workflows, enabling migration of session histories between those environments.
Does it protect data and fit developer workflows?
txcript adopts a local-first design so session conversions and searches run on the user's machine, and data only leaves when explicitly shared via team workspaces. The MCP server runs in read-only mode to expose past sessions as tools, for example through commands that list and read sessions. These behaviors make it suitable for developer and researcher environments that require local control of sensitive conversation history.
A practical tool for developers needing session continuity
txcript is a practical option for developers and researchers who require portable conversation continuity across multiple assistants; it preserves hidden session state and provides local search. Expect to treat converted transcripts as part of your toolchain rather than a drop-in replacement for every assistant, and validate critical sessions after migration when working on safety- or correctness-sensitive tasks.
