Saves technical teams significant time when switching between AI coding tools without losing context or starting tasks over.
What Is Skillsync And What Does It Do?
Skillsync is an open-source tool that moves AI coding agent chat sessions between agents, so a team can switch tools without losing the conversation. It transfers the entire session, including messages, reasoning, and tool calls, between Claude Code, Codex, and Cursor.
The core is txcript, an open-source Rust engine the founders describe as ffmpeg or pandoc for agent sessions. The launch post lays out the rest: a local-first Mac desktop app, a CLI, and MCP integration, with skills and memory stored as portable, human-readable markdown any agent can read.
Everything runs locally except what you explicitly share into team workspaces, which sync sessions so teammates continue each other’s work instead of repeating it. The product comes from skillsync.com, a YC W26 team of two founders, Nars and Nishant.
Skillsync turns disposable chat logs into a portable asset you own.
Is Skillsync Actually Maintained?
It is a day-one release, and the honest answer is that maintenance is early but measurable. Within hours of launch, a commenter flagged that archived Codex sessions don’t auto-discover, filed issue #52 on the txcript repo, and had PR #53 merged before the thread went cold.
Both founders answered technical questions in the launch thread through the day, and the core engine ships as an npm package other builders can pull in. One commenter already writing cross-harness integration tests called the shared translation layer the hardest part of their own problem.
The team’s track record is the strongest maintenance signal: their previous open-source project, a payment orchestrator that let merchants route across processors instead of getting locked into one, reached 30k+ GitHub stars.
A day-one release with a same-day fix cadence and a 30k-star prior project earns a trial, not trust.
Is Skillsync Good Enough For Production Handoffs?
For mid-task moves, yes, with one calibrated expectation. The flagship use case works: hit your usage limit on Claude Code, move the full session to Codex, and keep the task running with the transcript intact.
The founders are candid about the limits. With longer sessions, the receiving agent sometimes decides to compact it, though the whole transcript still exists and stays accessible.
Some harnesses obfuscate their reasoning, although the founders say almost everything goes through.
The alternative most teams use today, asking the agent to keep a markdown work journal, loses whatever the prompt didn’t ask about. A founder’s own explanation: the journal records one way of looking at the session, and you find out which detail you needed much later, while the raw session still has it.
Good enough for handoffs, not yet good enough to be your only copy of record.
The sprint stalled when the team’s primary agent hit its usage cap with the migration half-finished. Months of debugging context sat in one tool’s local format, readable by exactly one program, while the cheapest move was switching agents for the rest of the month.
Skillsync’s answer to that scene is a translation layer, not a cloud: txcript converts between the on-disk formats the way ffmpeg converts between video formats, and the session arrives in the second agent with its conversation, reasoning, and tool history mapped over.
The founders earned the benefit of the doubt on interoperability the hard way, taking a payments orchestrator past 30,000+ GitHub stars by fighting exactly this class of lock-in. The trial costs $0, and it runs on your machine.
What Breaks When You Run Skillsync In Production?
Three known rough edges from the launch thread itself. Long sessions sometimes get compacted by the receiving agent, archived Codex chats must be unarchived with the app restarted before they appear, and the flow is one-way today, from chat into your coding agent, not the reverse.
Syncing sessions across machines works through SSH from the desktop app, and the founders confirmed support for the Antigravity harness. Merging two colleagues’ sessions into one, which one commenter’s team handles with a homegrown workaround, is not something Skillsync does yet.
None of these are silent failures, which is the standard that matters. The transcript survives every one of them, and the engine is open source, so the failure modes are inspectable.
Day-one software with documented limits beats mature software with hidden ones, for exactly as long as the fix cadence holds.
Should You Install Skillsync Today?
The cost of trying is near zero: it is free, open-source, and local, so the trial risks an afternoon, not a budget. Install it, port one real mid-size session, and find out whether your context survives the move before you trust it with a critical handoff.
If your team runs a single agent and never bumps usage caps, the urgency is low, and your exit option just got cheaper anyway. If you run multiple agents or hand sessions between people, the portability layer is worth adopting this week.
The structural bet is sound regardless: an open-source translation engine survives even if the company doesn’t, which is the whole anti-lock-in thesis. We track the coding-agent infrastructure shift on the daily wire as it develops.
Install it today, test it on one session, and let the transcript tell you whether to keep it.
Source: Hacker News
