What happened
Google announced the shutdown in its October 29, 2026 Jules changelog entry, “Jules Turndown”. The company says it is winding down Jules as part of work to unify its agentic coding efforts. Jules will close on December 31, 2026.
There are two operational deadlines. Users have until November 11, 2026 to continue finishing sessions. After that date, Google will turn Jules into a read-only service until the end of the year. The notice does not describe a successor service, a feature-by-feature migration path, or any promise that an alternative Google product will reproduce Jules workflows. It says that lessons from Jules are shaping products across Google, including Google Antigravity, which it calls an agent-first platform.
Jules was introduced as an experiment in asynchronous coding agents: software that could work through a backlog rather than only answer in an interactive chat. Its changelog records features associated with that mode of work, including scheduled tasks, connected repositories, memories, setup scripts, sessions, a command-line companion, an API, MCP support, and the ability to create pull requests. The shutdown therefore affects more than source-code output. It can affect the operational context accumulated around ongoing coding work.
Google has published a local export path. Users install the Jules Tools CLI with npm install -g @google/jules, sign in, and use the supplied export prompt with a CLI coding agent. The export is intended to save account settings, environment variables, scheduled tasks, connected repositories, memories, setup scripts, and every session. For each session, the stated scope includes the prompt, plan, full activity log, and code patches.
The company distinguishes that exported Jules state from code-hosting artifacts. Code, branches, and pull requests already reside in GitHub and remain there. What must be preserved separately is the configuration and history held inside Jules. That distinction matters for migration: a repository may retain the final patch while losing the task setup, agent plan, execution record, recurring schedule, and environment context that made the work repeatable.
The published export instructions also make the sensitivity of that archive clear. Environment variables are among the exported materials, and the instruction explicitly says not to print their values. Teams should treat an export as an operational archive with secrets-handling requirements, not merely as a backup of developer notes.
What the archive adds
This is a new event on a previously covered product thread. In October, “unreal-agent: Async-first agent harness” framed long-running agent work as a runtime and persistence problem rather than a single prompt-and-tool loop. Jules’ closure makes that persistence question immediate for users of one vendor service: the work state now has to move or be archived.
Trend signal
The export reinforces the pattern that harness architecture outlives model choice. A coding agent’s durable value is not only the model that generated a patch; it is also the schedules, repository bindings, setup, memory, plans, logs, and controls around that model. Jules’ shutdown does not prove that every one of those elements can move cleanly between tools, but it shows why teams need to identify and retain them.
Why now
The timing is set by Google’s transition schedule. The announcement gives users until November 11 to finish sessions, then leaves a read-only period through December 31. That creates a short window to identify work that still depends on Jules being able to run, rather than merely needing its historical records.
Why it matters
An agent migration can fail even when all source code is safely in version control. Scheduled maintenance, repository-specific setup, environment configuration, task instructions, prior plans, and execution logs may live outside the repository. Those elements are often what let a team restart recurring work or diagnose why an agent changed a file.
For Jules users, the immediate job is to separate artifacts into three groups: code and pull requests already preserved in GitHub; service-held context that should be exported; and live workflows that need a replacement before the active-session deadline. The export offers a record, not evidence that another coding agent can consume it automatically. Rebuilding a workflow may require translating schedules, permissions, setup scripts, and review practices into a different runtime.
What happens next
- Teams with active Jules automations will complete or replace them before the active-session cutoff. (3 months, 80%). Watch for: migration guidance or user reports of rebuilt scheduled workflows.
- Jules will become read-only after November 11 and remain so until shutdown. (3 months, 90%). Watch for: a change in the Jules product interface or changelog confirming the access state.
- Google may clarify how Jules learnings appear in its wider agentic coding products. (6 months, 55%). Watch for: a product announcement that names Jules migration or equivalent capabilities.
Limits and uncertainty
Google’s notice establishes the dates, the read-only transition, and the stated export scope. It does not establish how many users or teams rely on Jules, whether the export is complete for every account configuration, or whether all exported records can be imported into another agent system. It also does not promise that Google Antigravity will replace Jules, preserve its API or scheduled-task behavior, or accept Jules exports. The announcement describes an export process but offers no independent evidence here about its performance on large histories or complex repository setups.
Skill to build now
Make an agent-state inventory this week. For each recurring or long-running coding workflow, record the repository, schedule, setup commands, required environment variables, permissions, task prompt, expected deliverable, and human review point. Then export Jules and store the result under access controls appropriate for secrets. Run one representative workflow with the prospective replacement agent and compare the resulting plan, patch, tests, and review trail. This turns a service shutdown into a testable migration rather than an archival exercise.
Reader takeaway
Jules users should not wait for the final shutdown date. Finish work that requires an active Jules session before November 11, export the service-held context, protect the resulting archive, and test replacement workflows against real recurring tasks. GitHub preserves code; it does not automatically preserve the agent operating system around that code.
Open question
Which parts of a Jules workflow can a team recreate from the export alone, and which will require manual redesign in a new coding-agent runtime?