Assumer CLI

The AI coding agent for your terminal

Assumer reads your codebase, plans the change, edits files, runs your tests and shows you the diff, all from the shell you already use.

npm i -g @4born-solutions/assumer

assumer — ~/projects/shop-api 
$ assumer
build · Add rate limiting to the /orders endpoint and cover it with tests

● read     src/routes/orders.ts
● grep     "express-rate-limit" in package.json
● todo     1. add limiter middleware  2. wire into /orders  3. tests
● edit     src/middleware/rate-limit.ts  +24
● edit     src/routes/orders.ts           +3 -1
● write    test/orders.rate-limit.test.ts +41
● bash     npm test -- orders
           ✓ 12 passed

Done. /orders now allows 60 requests per minute per API key and
returns 429 with a Retry-After header. Review the diff with /diff.
Example Assumer session: the agent reads the route, plans the work as a to-do list, edits two files, writes a test and runs the test suite.

Built for real codebases

Everything the agent needs to work on a production repository, with you in control of every step.

Understands your codebase

Reads, greps and globs across the repository, follows imports and uses language servers (LSP) to understand types, symbols and errors before it changes anything.

Plans before it edits

Switch to the read-only plan agent to analyse a problem and agree on an approach, then hand over to build to make the change.

Multi-file edits you can review

Edits, writes and patches files across the project, shows a diff for every change, and lets you /undo and /redo at any point.

Runs commands and tests

Executes shell commands, test suites and builds, reads the output and keeps iterating until the task is done, within the permissions you set.

Subagents for parallel work

Delegates research and exploration to subagents so the main session keeps a clean context on long tasks.

Granular permissions

Allow, ask or deny per tool and per shell command pattern, so the agent never runs anything you did not approve.

Commands

Run assumer for the interactive UI, or use these subcommands directly.

assumer Start the terminal UI in the current project.
assumer login Sign in to your Assumer account (device sign-in).
assumer run "<prompt>" Run a task non-interactively, for scripts and CI.
assumer models List the models available to your account.
assumer mcp add Connect a local or remote MCP server.
assumer agent create Create a custom agent with its own prompt, tools and permissions.
assumer github install Set up the GitHub integration for a repository.
assumer pr <number> Check out a pull request and work on it with the agent.
assumer meeting develop Turn a meeting transcript into a development plan.
assumer session list List, resume, export or import past sessions.
assumer stats Show token usage and cost statistics.
assumer serve Run Assumer as a headless server with an HTTP API.
assumer web Open the Assumer web interface on top of the server.
assumer acp Expose Assumer over the Agent Client Protocol for editors such as Zed.
assumer upgrade Update to the latest version.

Slash commands

Inside the terminal UI, type / to open the command palette.

/new

Start a new session

/sessions

Switch between sessions

/models

Pick a model

/agents

Switch agent (build, plan or custom)

/undo

Revert the last change

/redo

Re-apply a reverted change

/compact

Summarise the conversation to free context

/share

Share a session link

/export

Export the conversation

/mcps

Manage MCP servers

/skills

Browse available skills

/themes

Change the theme

/editor

Write a long prompt in your $EDITOR

/diff

Review pending changes

/status

Show session and account status

/help

List all commands and keybinds

Built-in tools

The agent works through a small set of well-defined tools, each covered by your permission rules.

bash

Run shell commands, tests and build scripts.

read

Read files, with line ranges for large files.

edit / write / patch

Exact string edits, new files and multi-file patches.

grep / glob / list

Fast search across the repository.

todowrite

Track multi-step plans as a live to-do list.

webfetch

Read documentation and web pages as context.

task

Launch a subagent for focused research.

skill

Load reusable SKILL.md instructions on demand.

Frequently asked questions

Is the Assumer CLI a terminal UI or a command?

Both. Run assumer for the interactive terminal UI, or assumer run "your task" to run a task non-interactively from scripts and CI.

Which operating systems are supported?

macOS, Windows and Linux on x64 and ARM. Install with npm, pnpm or bun.

Does Assumer change files without asking?

Only if you allow it. Permissions can be set to allow, ask or deny for each tool and for specific shell command patterns, and every change can be reverted with /undo.

Can I use my existing AGENTS.md or CLAUDE.md?

Yes. Assumer reads Assumer.md rules files and is compatible with AGENTS.md and CLAUDE.md, so existing project instructions keep working.

How do I connect external tools?

Add any Model Context Protocol server with assumer mcp add. Local (stdio) and remote (HTTP, with OAuth) servers are supported.

Start building with Assumer

Install the CLI in one command, sign in, and run your first task in under a minute.

npm i -g @4born-solutions/assumer

Download Assumer Read the quick start