xAI ships Memory in Grok Build — coding agent keeps project notes across sessions

xAI’s newsroom said Grok Build now has memory. As you work, it keeps notes on conventions, decisions, and project facts; later sessions read those notes before touching related code. The company frames it as Grok Build getting better the more you use it.

SOFTWARE desk — same-day primary from xAI that its coding agent now keeps project memory across sessions, a concrete step toward agents that retain team conventions instead of starting cold every run.

What the company says it does: Grok Build now carries conventions, decisions, and project facts from one session to the next. Notes are written in the background as you work and read back when you return to the project. File that carry / write-in-background / read-back picture as xAI’s. This desk did not sit on a later session.

Company framing on the same page: “Grok Build now has memory.” The post says later sessions read those notes before touching related code, and that “Grok Build gets better the more you use it.” File those lines as xAI’s. Do not file this as ChatGPT-style long-term personal memory for Grok chat. This post is about Grok Build — xAI’s coding agent / CLI-style build assistant — not the Grok chat product.

Demo pattern, still company: the page shows the agent storing a project convention — use `just test`, not bare `cargo test`, because the suite starts a test database first — in topic files, then recalling that convention on a later task. The same page shows Capture / Recall style topic notes under a workspace, including a testing topic. `just` here is a command runner that can start supporting services before the tests. File that just-test / topic-file / later-recall demo as xAI’s. This desk did not run the suite.

How capture and recall work, as the same post has it: after a turn completes, Grok reviews it in the background and records anything durable — conventions, decisions, and project facts. Capture runs on every completed turn and does not interrupt the session. Notes are markdown files, one topic per subject. Each project has its own workspace scope; a global scope holds preferences that apply everywhere. `/dream` merges new observations into the topic they belong to, and `/memory` shows the files as they are on disk. Before starting related work, Grok reads the topics that cover the area and applies them, including in sessions where the subject never comes up. Instructions in the current conversation take precedence over anything in a note. File that capture / markdown / dream / memory / current-conversation-wins picture as xAI’s.

What it remembers versus leaves out, company: memory holds how the team writes and reviews code, decisions and the reasoning behind them, and durable facts about the project — from where a subsystem lives to which command runs the suite. Task state, tentative conclusions, secrets, and anything the repository or its docs already cover are left out. File that remember / leave-out split as xAI’s. This desk did not inspect a memory file.

Availability, company: “Memory is available in Grok Build now.” It applies to new sessions: run `/new` or start a fresh `grok`, and notes begin after the first completed turn. File that available-now / new-sessions line as xAI’s. This desk did not install the CLI or start a session. Do not file that memory is on by default for Grok chat.

Install / try language on the same newsroom page, company CTA only: the page includes `curl -fsSL https://x.ai/cli/install.sh | bash` and a “Try Free” control. File those as company call-to-action copy — not an independent availability audit, not a price, and not a claim this desk installed the tool.

Plain English for the rest of the card: Grok Build = xAI’s coding agent / CLI-style build assistant. memory here = durable project notes the agent writes and reloads across sessions so it does not relearn the same repo rules every time. convention = a house rule for how this repo is supposed to be worked, such as which command runs the tests. session = one run of the agent on a project. topic file = a markdown note about one subject, such as testing. `/memory` = the company command that opens those note files. `/dream` = the company command that folds new notes into those topic files. CTA = call to action, the company’s own install / try copy.

PRIMARY here: xAI’s 16 Sep 2026 company newsroom — Tier A PRIMARY company source, the original record. The x.ai/cli product page is product-home context, not a second originating newsroom. The Memory in Grok Build post, the conventions / decisions / project-facts carry, the write-in-background / read-back picture, the “Grok Build now has memory” / “gets better the more you use it” framing, the just-test / topic-file / Capture-Recall demo, the capture / markdown / `/dream` / `/memory` / current-conversation-wins mechanics, the remember / leave-out split, the available-now / new-sessions line, and the curl / Try Free CTA are company-attributed. Product behavior stays company-attributed — not independently audited here. NOT claimed: pricing, usage metrics, eval scores, that this is ChatGPT-style long-term personal memory for Grok chat, that memory is on by default for Grok chat, that this desk installed Grok Build or watched a later session, a stock tip, or investment advice. Distinct from the already-filed factory-200m-5b-valuation, coder-claude-code-agent-relay, bolt-forge-open-models, stackhawk-wingman, typesafe-ai-40m-seed, and galileo-jupiter.