Private beta · local-first by default

One local command center for every coding CLI.

SupraSpace turns a tangle of terminal windows into one project command center — launch, watch, and steer multiple coding CLIs with live terminals, tasks, durable Markdown memory, and clear process status.

Join the waitlist See what it does

Local SQLite + project files · zero core telemetry · configurable profiles for Claude Code, Codex, OpenCode, Kiro CLI, and command-based CLIs

supraspace.app

api-gateway

Search projects, tasks, sessions…

Dashboard

4 active sessions

Backlog

Ready

Running

Blocked

Review

126

Done

Refactor auth router

Refactoring Specialist · Claude Code

Running

Elapsed 04:12

Last out 0:03

✓ extracted middleware chain
› applying edit 2/3 …

OpenStopReviewDone

Add rate limiting

Code Reviewer · Codex

Waiting

Elapsed 01:48

Last out 1:12

? Apply suggested limiter to /login? (y/n)

OpenStopReviewDone

Migrate to Postgres 16

Postgres Pro · Custom

Idle

Elapsed 12:30

Last out 3:40

› EXPLAIN ANALYZE complete · 0 seq scans

OpenStopReviewDone

Board

api-gateway · 18 tasks

Backlog3

Document webhook payloads

Technical Writer

Audit dependency tree

Security Auditor

Ready2

Add OpenAPI schema

API Designer

Running4

Refactor auth router

Refactoring Spec.

Migrate to PG 16

Postgres Pro

Blocked1

Wire payment webhook

Payment Integ.

Review2

Add rate limiting

Code Reviewer

Done6

Cache layer for /search

Performance Eng.

Sessions

2 × 2 grid · 4 live

Running

Refactor auth router

Refactoring Specialist · Claude Code

api ❯ claude --agent refactorer

› loading .supraspace/ memory …

✓ context loaded · 6 files

› analyzing router.ts

Waiting

Add rate limiting

Code Reviewer · Codex

api ❯ codex review --diff

✓ scanned 14 files

⚠ suggest limiter on /login

? apply fix? (y/n)

Running

Migrate to Postgres 16

Postgres Pro · Custom

api ❯ psql -f migrate.sql

› ALTER TABLE sessions …

✓ 4 migrations applied

› reindexing

Running

Write integration tests

Test Automator · OpenCode

api ❯ opencode --agent test

› generating cases …

✓ 42 passing

› coverage 91%

Memory

space map · 6 files

Project Memory

architecture.md

decisions.md

agent-notes.md

tasks.md

Fastify + Drizzle

SQLite store

node-pty / ConPTY

argv, not shell

append-only logs

MCP & Integrations

connected

MCP servers

filesystemstdio · local

ready

fetchhttp · remote

ready

postgresstdio · local

ready

Integrations

GitHubSource control

connected

VercelHosting

connected

CloudflareInfrastructure

connected

Swarms

ordered pipeline

01 · Scope

Claude Code

Done

Handoff: route map and acceptance notes saved.

02 · Implement

Codex

Running

Receives the prior step’s output and project context.

03 · Verify

OpenCode

Manual

Advance by process completion, idle detection, or your call.

Bring your own CLI

Profiles for the CLIs you run today.

Reusable profiles cover explicit Claude Code, Codex, OpenCode, and Kiro CLI support, plus configurable command-based CLIs. Choose the command, prompt delivery, and context each profile needs.

Claude Code

claude --agent <profile>

Anthropic's agentic CLI runs natively — your task, memory, and skills load on boot.

Codex

codex --prompt <task>

OpenAI's coding CLI plugs straight in with clean prompt-flag injection.

OpenCode

opencode --agent <profile>

First-class support, including agent import from your opencode.json.

Kiro CLI + custom

your-tool {{prompt}}

Use the Kiro CLI profile or define a command template for the interactive CLI already in your toolbox.

Configure how a profile receives the work — stdin, args, file context, or a prompt flag — then reuse it across projects.

Drives any CLI · plugs into your stack

One workspace, many processes

Built for running many agents, not babysitting one.

A control room for parallel sessions: a board that mirrors real progress, project memory that stays in files, and status rooted in what the process is actually doing.

Run agents in parallel

Launch a whole crew at once and watch them in a 1, 2, 4, or 6-up terminal grid. Each agent gets its own task, profile, and live PTY — no more juggling a dozen terminal tabs.

Status you can trust

No spinner that lies. Status comes from real process liveness, output recency, and your own patterns — so running means running and waiting means it needs you.

Lightweight by design

The core workflow keeps its SQLite data and project files local. It is built to keep the command center close to the code and its working context.

Memory the agents read

Persistent project Markdown captures architecture, decisions, and notes. It can be included in launches and written back, so useful context stays beside the code instead of disappearing into a chat.

An extensibility layer

Configure MCP servers, skills, hooks, tools, presets, and bindings for the way you work. Starter and marketplace MCP entries may still need their own setup.

Optional remote control

Use the local command center at your desk, or deliberately configure a remote listener with TLS for restrained mobile control. Remote access is off until you configure it.

SupraSpace

Swarms

Ordered pipelines, with handoffs you can inspect.

Put steps in order

A swarm is an ordered pipeline. Give each step its own profile and task, from discovery through implementation and verification.

Carry work forward

Each completed step hands its output to the next one, creating a visible chain rather than disconnected sessions.

Advance on your terms

Move a swarm by process completion, when a process becomes idle, or manually when you want to review a handoff first.

Kanban board

A board that tells the truth.

Drag a task and its agent follows. When a session starts the card flips to running on its own; when an agent finishes, it lands in review. The board is a live mirror of what your agents are actually doing.

Drag-and-drop across backlog → ready → running → review → done
Quick-launch an agent straight from a card
Status updates flow back from real session lifecycle

Running

Refactor auth router

Migrate to PG 16

Review

Add rate limiting

Done

Cache /search

Fix flaky test

Live terminals

Real terminals, streamed live.

Every session is a genuine PTY streamed to your browser with xterm.js. Read-only previews keep the grid calm; pop any session out for a full interactive terminal with scrollback, keyboard, and links.

1 / 2 / 4 / 6-panel layouts that keep live connections
Type back the moment an agent waits for input
Append-only transcripts survive every restart

api ❯ claude --agent refactorer

› loading project memory from .supraspace/ …

✓ context loaded · 6 files · 2 decisions

› extracting middleware chain

✓ refactor plan ready · 3 edits

› applying edit 1/3

Project memory

Context that compounds.

Project Markdown captures architecture, decisions, and agent notes alongside your code. You can choose what launches receive and write useful outcomes back, with an explorable map of the project context.

Transparent Markdown — no hidden vector black box
Pick which snippets ride along on each launch, then write back what matters
Interactive space-map view of the whole project

Memory

architecture.md

decisions.md

agent-notes.md

tasks.md

Privacy & security

Local by default. Explicit when it reaches out.

Loopback first. The default listener binds locally; core workflow data lives in local SQLite and project files with zero core telemetry.

Your configuration changes the boundary. Configured LLM providers, integrations, remote MCP, web push, and optional remote access can use the network.

Remote is deliberate. Mobile control is optional and requires an explicitly configured remote listener and TLS.

6-up

Terminals at a glance

Multi

Reusable CLI profiles

Local

SQLite and project files

PTY

Live terminal sessions

Connect your world

Configure the connections your workflow needs.

Integrations are configured by you. When you connect a provider, service, or remote MCP server, it can make network requests as part of that configuration.

LLM providersYour configuration

IntegrationsYour configuration

Remote MCPOptional network

SkillsReusable context

HooksConfigured actions

BindingsWorkflow wiring

PresetsReusable setup

OpenAIAI

SlackNotifications

MCP serversLocal & remote

Private beta

Get early access.

SupraSpace isn't public yet. Join the waitlist for early-access updates.

Your email is sent to Formspree to manage the waitlist. We use it only for SupraSpace early-access updates.