Shipped is an Atlas command that answers what did you get done this week from your actual history rather than your memory. Run /shipped and Atlas reads your git commits and working sessions for the last 7 days, groups them, and writes a plain-language document to narrative/<window>.md in your repository. It is free, on by default, and needs no account. Four further sections fill in when Atlas also has a record of what was said around the work.
What did you get done this week
The record of a week already exists across 2 places, your commits and your Atlas sessions, and neither one reads like a sentence.
Shipped gathers both for a time window and renders them as prose rather than a commit list. There is nothing new to log, because it reads what is already there. The default window is 7 days, and --since changes it.
It writes markdown into your repository
Shipped writes 1 file per window, narrative/<window>.md, as ordinary git-tracked markdown.
The document is yours and portable: readable in any editor, diffable in review, and regenerated per window rather than maintained by hand. Atlas keeps a rebuildable cache alongside it, but the markdown is the artefact you keep.
The sections it cannot fill, it names
Shipped renders 1 section from git alone, and names the 4 it cannot fill without more context.
Shipped comes from commits and sessions, so it always renders. Decided, Promised, Learned, and Scope changes come from meeting transcripts read by Atlas Intake. When Intake is not connected, the document lists those four by name under a Not recorded heading rather than silently returning a shorter file.
Availability
Shipped is the 1 record engine included in the free Atlas core, and it is available today.
It defaults on with no configuration and requires no account, though it does need a model provider configured, which Atlas needs anyway. The Intake-backed sections belong to the paid tier, and Atlas Pro is not yet on sale, so those four sections are not reachable yet.
Frequently asked questions
- Is the Atlas work log available now?
- Yes. Run /shipped in Atlas. It is part of the free core, on by default, and requires no account.
- What does it read to build the log?
- Your git commits and your Atlas working sessions for the window. If Intake is connected it also reads the commitments, decisions, facts, and scope changes extracted from your meeting transcripts.
- Do I have to write anything?
- No. The log is derived from the record Atlas already has and is regenerated per window, so there is nothing to maintain by hand.
- Where does the file go?
- narrative/<window>.md inside your repository, as ordinary git-tracked markdown. It is yours, and it stays readable and diffable without Atlas.
- Why does my log say Not recorded?
- Because Atlas has no meeting record for that window. Shipped is built from commits and sessions, while Decided, Promised, Learned, and Scope changes need Intake, which reads call transcripts and is part of the paid tier.
- Can it help with client status updates?
- It gives you the factual half: what actually shipped, in sentences. Shaping that into a client-ready note in a specific voice is a separate paid capability that is not yet on sale.
Try Atlas in your terminal
The terminal-native AI coding agent. Free core, single binary.
Install AtlasRelated guides
Atlas with Code Llama (local via Ollama): a Fill-in-the-Middle Baseline in 2026
Code Llama runs free and local via Ollama at 7b, 13b, 34b, and 70b with a 16,384 token context. Strong at infilling, but too small and too old for Atlas agent work.
Atlas with Qwen2.5-Coder 7B (Ollama): the default fully offline setup for 2026
Qwen2.5-Coder 7B (Ollama) in Atlas: the 4.7GB default tag with 32K tokens (32,768) of context, Free (self-hosted), and no cloud key anywhere in the agent loop.
Atlas with Gemini 2.5 Flash: The Workhorse small_model Pick for 2026
Gemini 2.5 Flash in Atlas: reasoning enabled, a 1,048,576 token context, and $0.3 per Mtok input, a sixth of what a frontier tier model charges to read the same repo.
Atlas with DeepSeek Reasoner: Chain-of-Thought Debugging at $0.28 in 2026
DeepSeek Reasoner in Atlas: visible reasoning traces on a 1M window at $0.14 / $0.28 per Mtok, roughly 640x below GPT-5.5 Pro's $180 output rate. Setup and limits.
Atlas with Llama 3.1 8B (Ollama): 128K Context on an 8GB Card in 2026
Run Atlas on Llama 3.1 8B (Ollama): Meta's 4.9GB workhorse with a 128K context, free self-hosted. Setup, the KV cache catch, and when a coder model wins.
Atlas with IBM Granite 3.3 8B (Ollama): the Free Local small_model for 2026
IBM Granite 3.3 8B (Ollama) is a 4.9GB general model with 128K tokens (131,072) of context, Free (self-hosted). Assign it to small_model in Atlas. Setup and limits.
Atlas with CodeGemma 7B (Ollama): Fill-in-the-Middle on 8GB in 2026
CodeGemma 7B (Ollama) is Google's 5.0GB code model with fill-in-the-middle training and an 8K context, Free (self-hosted). Atlas setup and honest tradeoffs for 2026.
Atlas vs Blackbox AI: Choosing Your AI Coding Agent in 2026
Comparing Atlas, the terminal-native AI coding agent, with Blackbox AI, a VS Code agent with 4.7 million installs, for developers in 2026. Evaluate features, pricing, and workflow.