Skip to content
Beta
Sahabti
Last updated · August 30, 2026
v0.4.0-beta

What's new in Sahabti.

Every meaningful change to the platform, in reverse-chronological order. Subscribe to the RSS feed and we'll keep you posted.

Subscribe to RSSBrowse history
v0.4.0-betaFeature

Deploy from GitHub, database tools, and an MCP endpoint

Import a repository and ship it to a live URL on every push, browse and query PostgreSQL and MySQL from the dashboard, connect an AI agent over MCP, and scope API tokens down to single actions.

Deploy from GitHub, database tools, and an MCP endpoint

Until now Sahabti ran your data. This release runs your code too.

What's new

  • Projects — deploy from GitHub. Connect GitHub, import a repository, and Sahabti detects the stack, builds a container, and serves it over HTTPS at a hostname of its own. Every push to the tracked branch ships a new version; a new container has to pass a health check before it takes over, and a failed one never reaches your visitors. Roll back and the exact image that was serving before goes live again — no rebuild. See the projects docs.
  • Database tools. PostgreSQL and MySQL instances now carry a table editor, a SQL editor with history and CSV/JSON export, and a database users panel for handing each app or teammate its own account. Queries run inside the instance rather than over a connection from your browser, so a private instance is fully usable from the dashboard. See database tools.
  • MCP for AI agents. https://api.sahabti.com/mcp puts twenty-six tools in front of Claude, Cursor, or your own agent — list and launch instances, read a failed provisioning log, run a query, take a backup, request more quota. Sign in with OAuth, or hand an unattended agent a scoped token. Connection secrets are deliberately not exposed to any tool. See MCP.
  • Least-privilege API tokens. Minting a token now means picking exactly the resources and actions it needs — instance:read without instance:delete, lifecycle without query — capped at your own role, with an expiry up to a year. See API tokens.
  • A workspace notification feed. What the platform did while you were away: instances that came up, the ones that failed and on which step, deployments that shipped.
  • Retry a failed provision. An instance that stopped on an infrastructure step can be picked up again over the resources it already holds, instead of being deleted and recreated.
  • Per-size network caps. Each instance size now carries a published network rate, so one busy neighbour can't take the uplink with it.

Everything here is live in the dashboard today.

v0.3.0-betaFeature

Public IPs your way, two more engines, and on-demand backups

Give an instance its own dedicated public IP or share the workspace's, launch FerretDB and Qdrant alongside the six existing engines, back up PostgreSQL on demand, and resize a running instance live.

Public IPs your way, two more engines, and on-demand backups

Everything here is live in the dashboard today.

What's new

  • Public IPs, your way. When you make an instance public, pick how it's reached: a dedicated IP — its own floating address on the engine's native port — or a shared IP that rides the workspace's ingress address on a per-instance port. Dedicated is a paid-plan option; the Free plan runs on a shared IP.
  • Two more engines. FerretDB (MongoDB-wire, on 27017) and Qdrant (vector search, on 6333) join the catalog — eight engines you can launch now, up from six.
  • On-demand backups. Take an engine-native logical dump of a running PostgreSQL instance from its Backups panel, then restore it into a fresh instance without ever touching the source. PostgreSQL is the first engine in the panel; more join as their dump scripts ship.
  • Live resize. Move a running instance to a bigger size from its Settings tab — the new CPU and memory apply live, no restart — as long as the new footprint fits your quota. Disk stays as provisioned.

See the instances and backups docs for the full picture.

v0.2.0-betaFeature

Quota-based billing, live provisioning logs, and API tokens

A shared pool of vCPU, memory, and disk instead of per-service bills, a real-time provisioning timeline, live instance metrics, and scoped tokens for automation.

Quota-based billing, live provisioning logs, and API tokens

Three things that make running more than one service on Sahabti easier to manage.

What's new

  • Quota-based billing. Your plan grants a shared pool of vCPU, memory, and disk — spin up as many instances as fit, instead of paying per service. Start free, upgrade when you outgrow it.
  • Live provisioning timeline. Watch each step — network setup, container creation, engine install, health check — stream in on the instance page as it happens, instead of a spinner.
  • Live instance metrics. CPU, memory, disk, and network read straight off the hypervisor, over whichever window you pick — no agent to install.
  • Scoped API tokens. Create a workspace token to drive instances from a script or CI, and revoke it from the same page the moment you're done with it.

Read how to wire up a token in the API docs.

v0.1.0-betaFeature

Managed services, isolated by design

PostgreSQL, MySQL, MariaDB, Valkey, Memcached, and RabbitMQ — each in its own private network, with a public endpoint, ready in about 75 seconds.

Managed services, isolated by design

Sahabti is live: pick an engine from the catalog and get a running, isolated instance in about 75 seconds — no networking, TLS, or firewall config to write yourself.

What's new

  • Six engines at launch. PostgreSQL 17, MySQL 8.4, MariaDB 11, Valkey 8, Memcached, and RabbitMQ. The rest of the catalog is listed in the dashboard and marked coming soon until its provisioning script ships.
  • Isolated by design. Every instance runs in an unprivileged container on its workspace's own private network. Tenants can never reach each other — isolation is enforced at the network layer, not just a password.
  • Reachable from anywhere. Each instance gets a public hostname and port that proxies straight into your private network, and the SQL engines come up with the platform certificate already installed.
  • Fast provisioning. Pick an engine and a size; the network, container, and endpoint are provisioned automatically, and connection details are ready the moment it's up.

Launch your first instance from the dashboard — the catalog, sizing, and connection details are all one page.