Visual workflows
Place actions, storage endpoints and flow control (fan-out, loops, decisions) on a canvas with auto-layout, snap to grid and undo.
Beam Studio is a self-hosted workspace for Beam. Build transfer workflows on a visual canvas, run them by hand, on a schedule or from a webhook, and connect machines and buckets in private rooms.
Triggers, storage endpoints and actions on one canvas.
Machines, buckets and Studio, joined by invitation.
Signed releases with health checks and rollback.
Room connections
9 members · 36 permitted connections
Lines show what each member may send, not live traffic.
Start from a blank canvas or the Simple transfer template. Connect a trigger to storage endpoints and a Beam Transfer step, add actions after it, and check the credit estimate before you run.
A room is a private group you join by invitation. Channel permissions decide which member may send to which, and Studio draws them for every room.
Agents on your machines, buckets from your storage credentials, and the Studio room consumer. A machine enrolls with a one-time code over an outbound connection, so it needs no inbound port.
Message, stream, datagram, request-reply, media and object channels. Each grant says who may send and who may receive, and the graph draws exactly that.
Share a room with a role, channel permissions, an expiry and up to 512 uses. The recipient joins from the Beam CLI:
beam room join <room> --invitation-token <token>Objects sent between agents are end-to-end encrypted with the room's keys, so workers carry ciphertext only. Bucket members use TLS.
Studio adds composition, scheduling and visibility around Beam transfers. Beam still runs every transfer, and each one appears in your Console.
Place actions, storage endpoints and flow control (fan-out, loops, decisions) on a canvas with auto-layout, snap to grid and undo.
Run by hand, on a schedule, from a webhook that can require a signed body, at a set time, or after another workflow.
Beam Transfer comes preinstalled. Add Slack, HTTP, webhook, Salesforce and Zapier steps, plus public and private Registry actions.
Amazon S3, Cloudflare R2, Hippius, Hugging Face, Wasabi, Backblaze B2 and more, plus any S3-compatible endpoint. Credentials stay encrypted on your host.
22 tools let AI agents build, run and inspect workflows and transfers, using scoped, expiring and audited tokens.
Chat with Beam AI to draft workflows and plan operations, billed to your organization's credits.
Every run keeps its graph, step states, retries and outcome. Credits are reserved before a run dispatches and settled when it ends.
Studio runs on one Linux host with systemd and Docker Compose, on amd64 or arm64. Plan for 2 vCPUs, 4 GB of memory and 20 GB of disk.
curl -fsSL https://cdn.b1m.ai/studio/install.sh | sudo shInstalls Docker if it is missing, then serves Studio at http://localhost:3004. Sign in with your Beam account.
One command sets up the signed stack and the beam-updater service. Re-running it keeps your settings.
A new install serves no one until its owner claims it with the code from sudo beam-updater claim-code.
Releases are Ed25519-signed and pin every image by digest. An admin confirms each update from Settings.
Every update backs up the database first and returns to the previous release when a health check fails.
Install it on a Linux host, sign in with your Beam account, claim it, and build your first workflow.