The command center for AI agents.

Sessions

Sessions in tabs

As many sessions as you need: each agent brings its own quick commands, the status sits on the tab itself, and the browser, PR checks and the emulator open as tabs of the same kind.

Three session views

Native markdown render with tool chips, thinking blocks, token and cost counters — or the raw PTY, or both.

Session persistence

History matched against the agent's own sessions, so a restart picks up where you left off.

Command palette

Fuzzy search over actions, terminals, recent sessions, projects and workflows — launch anything without leaving the keyboard.

Model, effort, queue

The model and the reasoning effort are picked right in the prompt field, and messages can be queued while the agent is still busy.

Activity feed

Everything the sessions do, in one stream on the home screen, with filters for prompts, tools, errors and finished runs.

Floating widget

An overlay above other windows: agent status at a glance, permission prompts, one click back into a session.

Automation

Tasks & kanban

Markdown descriptions, image attachments, branch targeting, worktree isolation. Start a task and it launches an agent, handing it the description as the prompt.

Visual workflow editor

A graph with parallel branches, conditionals, manual approvals and connector actions.

Headless execution

Agents and scripts run in the background and write a full log you can read back later.

Built-in MCP server

Agents create tasks, trigger workflows, spawn sessions, read terminal output and ask you questions — nothing to set up.

Connectors

Issues and PRs from GitHub and tasks from Linear become ordinary tasks inside BaseGrid IDE, and a new one can start a workflow by itself.

Code

Changes & inline review

Stats for every file, inline diffs, click a line to comment — all comments reach the agent as one structured message.

PR checks

CI statuses and reviewer comments next to the session, one keystroke away from GitHub itself.

Git & worktrees

Creation, auto-setup, merge menu, safe removal. Commit messages can be written by an agent.

Notes & file tree

Worktree-scoped markdown notes, and file previews with syntax highlighting — spreadsheets open as spreadsheets.

Around it

basegrid.json

Project config in the repo: run scripts, agent settings, worktree rules, env, task and workflow templates, MCP servers. Hot reload.

Remote hosts

Terminals and sessions over SSH, with keys in the built-in vault.

Your phone

The same local server hands out a PWA — scan the QR from settings. From another network, a wizard sets up your own VPS with a reverse SSH tunnel and HTTPS. A native Expo client lives in the repo.

Offline voice dictation

Local speech recognition via sherpa-onnx — hold space in the prompt field to dictate.

Mobile emulator

An Android device over scrcpy or an iOS simulator, live in a tab. You tap and type on it; the agent drives the same device through MCP, so you both see one state.

Terminal dock

Shell tabs in the sidebar, plus a run stack for the scripts declared in basegrid.json and their output.

Workspaces

Projects grouped into workspaces — one switch moves the whole context.

Built-in browser

A browser panel with console and network requests, one tab over from the agent.

Status bar

Claude usage limits, CPU and memory, active sessions, busy ports — a process can be killed right from there.