A bold move towards CLI

We have rebuilt Byterover from the ground up.

Over the past few months, we’ve launched two major versions - Byterover 1.0, built entirely on MCP, and Byterover 2.0, introducing Git for context.

The previous two versions truly brings significant values about context management: we have seen teams save thousands of dollars in LLM tokens every month, and how Byterover has become an indispensable workspace to unify context for agents.

However, through continuous feedback from our users and testing, we learned that MCP simply was not the best way to bring context engineering directly into agentic coding. MCP installation and tool triggers of previous version were not consistent across operating systems, IDEs, extensions, coding CLIs, or LLMs.

Therefore, we made a bold move - to build a CLI-native version of Byterover.

As a developer-first team, and as context becomes increasingly important to agentic coding, we want to give developers the most reliable way to work with memory/context.

⚠️ A side note:

This version is still in beta. We recommend trying it first on personal or side projects.

👉 Sign up for the Beta now →

👉 Read documentation →

One-line install, anywhere

Byterover CLI can now run consistently in any OS (macOS, Linux, or Windows), and works seamlessly across coding models and agentic coding platforms (20+ including Claude, Cursor, Codex, and more). Because context commands are CLI-native, even low-tier models can manage memory effectively.

npm install -g

or check the quickstart guide → in the doc.

Memory structure revamp and more human-in-the-loop control.

We have also redesigned the structure of memory and upgraded how ByteRover stores and retrieves memory.

• The new structure of the memory in the local space (.brv/ace/playbook.json) will capture context as structured “bullets” that can be searched and reused.

• You will have more control over how context is chunked or stored and how it gets retrieved.

• You can use the new Agentic Context Engineering (ACE) workflow that let agents complete tasks and auto-update local memory.

• You can now review and approve local context before they’re pushed to the workspace on Byterover webapp. When the local context is pushed to ByteRover, it’s organized into a knowledge graph that captures both semantic relationships and code structure.And many more.

Core commands for memory operations

Store or update markdown context in your local space. Read more about adding/updating memory in the documentation here.

Run the ACE Workflow - Agentic Context Engineering - the new workflow that allows agents complete tasks and auto-update local memory. Read more about ACE here.

Push local context to your remote workspace on Byterover web app.

Query relevant context from your workspace to feed your agent. Read more about retrieving memory in the doc here.

We are still keeping Git for context as our core

(1) Make agent able to use only exact piece of context for every coding task, helping teams improve code quality by AI, and saving tons of tokens.

(2) A reliable place for teams to share, manage and version context.

Just from now on, the way you interact with Byterover is different, which is directly on your terminal.

Get started now

👉 Sign up for the Beta now →

👉 Read documentation →

⚠️ A side note:

This version is still in beta. We recommend trying it first on personal or side projects.

We hope that this CLI version will unlock a new effective way for developer teams to work with context. Please share with us your feedback as you try it out.

• 💬 Email: andy@byterover.dev

• 📅 Book a quick call: cal.com/andy-nguyen/15min