djobs

Token-saving durable context for your AI coding agent. djobs gives Codex, Claude, Gemini, Copilot, Cursor, Cline, and any MCP-compatible agent durable, resumable task memory so long, multi-file work survives IDE crashes, context loss, and closed chats - nothing lost, nothing redone, fewer tokens replaying finished work.

This extension is the one-click way to set djobs up: it installs and manages the runtime for you, wires the MCP server, installs the agent instructions, and adds a task sidebar. You do not manage Python or config by hand.

Get started

1. Install this extension.
2. Run djobs: Set up / Repair djobs from the Command Palette (or click Set up djobs when the extension offers).
3. Start a new agent session and keep talking normally.

The extension does not generate, copy, or open Chat prompts. It wires MCP tools, installs agent guidance, and shows durable task state in the sidebar. A compatible MCP agent can then call resume_session, create djobs tasks before multi-step edits, and finish each unit with evidence.

After that, keep talking normally: "continue", "fix this", "run tests", "retry", "the previous run failed", or "release" are enough. The installed agent guidance tells the AI to bring djobs in before editing; you do not need to mention djobs in every prompt.

The setup step installs the runtime if needed, registers the MCP server, installs the agent guidance block, and verifies everything with djobs doctor. Use djobs: Diagnose Setup anytime to re-check.

Compatibility status

This extension is implemented and tested with GitHub Copilot in VS Code. djobs is MCP-based and is intended to work with Codex, Claude, Gemini, Cursor, Cline, and other MCP-capable coding-agent hosts, but those non-Copilot flows are not yet fully end-to-end tested.

Upgrading from an older version

Version 0.8.5 removes the older prompt-driving commands and auto-takeover settings. After upgrade, the extension only manages setup, MCP registration, diagnostics, and the task sidebar. It also checks the VS Code Marketplace at most once per day and tells you when a newer djobs extension version is available, because VS Code auto-update can lag or be disabled.

Version 0.8.6 adds visible Enable Prompt Actions and Disable Prompt Actions commands to the sidebar toolbar and Command Palette.

Version 0.9.0 adds Pause djobs / Resume djobs and updates stale Python packages automatically when the installed CLI is too old for those commands. It also pairs with the new AI Work Receipt (djobs receipt / the work_receipt MCP tool), an evidence-backed, git-aware summary of what the agent actually did.

Prompt actions are available only if you explicitly enable djobs.promptActions.enabled. When enabled, the sidebar shows Prompt Agent to Finish Workflow; when disabled (the default), djobs never opens Chat or asks whether to enable prompt actions. Use the sidebar toolbar command Enable Prompt Actions (play-circle icon) or the Command Palette command djobs: Enable Prompt Actions to turn it on for the current workspace.

What the sidebar shows

• djobs tasks grouped by workflow and status.
• Explicit djobs: Active/Paused and Prompt actions: On/Off rows, so you can see switch state without guessing from toolbar icons.
• A Pause djobs / Resume djobs toolbar button. Pausing temporarily stops agents from resuming or enqueueing tasks (useful when a stuck task keeps dragging the agent back); it deletes nothing and is reversible.
• Stale (long-unfinished) and blocked tasks flagged, with one-click archive.
• Right-click a task to archive it, permanently delete it, view its audit history, copy its ID, or inspect its raw JSON.
• Skip or archive stale work without asking an agent to do it. Archive keeps the audit trail; delete removes the task and its events.
• When the current workspace has no active tasks, the empty state explains that tasks appear when an MCP-enabled agent records them.

Advanced

The extension drives the djobs CLI under the hood. If you prefer to manage the runtime yourself, install it once globally (uv tool install djobs or pipx install djobs with Python 3.11+) and run djobs init in a project; the extension will detect and use it. If djobs lives in a non-default interpreter, set djobs.pythonPath in VS Code settings.