SpringRoll

From Springboard to Rollout

Sign inSign upDeploy now

SpringRoll setup guide

Ship your first app.

Use the agent you already work with. Connect it to SpringRoll, deploy a development preview, then ask a person to approve production and data access.

How it works

Your agent builds it. SpringRoll takes it from there.

SpringRoll does not replace Claude Code, Codex, ChatGPT, or your repository. It gives the finished app an owner, a release history, three environments, approvals, and a controlled route to company data.

01

SpringRoll Deploy

Move agent-built software from an untracked project into a governed production record.

  1. Coding agents

    Claude Code · Codex · ChatGPT · Claude Desktop · compatible MCP clients

  2. App Registry

    Owner · classification · immutable release · provenance

  3. DEV → UAT → PRODUCTION

    Environment-specific deployments, promotion and rollback

  4. Human gate

    Approval · RBAC · policy evidence

  5. Governed runtime

    Vercel today, behind the portable RuntimeProvider interface

Coding agents then App Registry then DEV → UAT → PRODUCTION then Human gate then Governed runtime.

02

SpringRoll Connect

Give each application its own identity and only the enterprise data it was approved to use.

  1. Enterprise source

    PostgreSQL · Notion · Salesforce · SAP

  2. Connector

    SpringRoll retains the source credential

  3. Data Product Catalog

    Named, discoverable, governed datasets

  4. Access approval

    App-specific, expiring and revocable grants

  5. Data Gateway

    Identity · allowlists · filters · masking · limits · logs

  6. Approved applications

    Finance and Demand receive independently governed views

Enterprise source then Connector then Data Product Catalog then Access approval then Data Gateway then Approved applications.

Credential boundary: applications never receive the source-system credential.

Connect an agent

Connect once, then work from your project.

Choose your client and copy the command. Your browser opens so you can choose an organization and approve the connection. You never paste or store an access token.

Terminal. One command, then authorize.

  1. 01

    Add SpringRoll

    Run this once, from anywhere.
    claude mcp add --transport http springroll https://springroll.dev/api/mcp
    Then start Claude and authorize. Your browser opens by itself.
    /mcp
  2. 02

    Authorize

    Choose your organization in the browser and approve the connection. No credential is exposed for you to store.

  3. 03

    Deploy

    Ask the agent to register, validate, and deploy the project.

    /springroll deploy

Describe the app

Tell SpringRoll what it is shipping.

The manifest names the app, its owner, classification, source, and framework. SpringRoll keeps hosting details out of this file, so the app record is not tied to one runtime.

apiVersion: springboard.dev/v1alpha1
kind: Application
metadata:
  name: demand-planner
  owner: supply-chain@acme.example
  department: Operations
spec:
  classification: CONFIDENTIAL
  source:
    git: github.com/acme/demand-planner
  framework: nextjs

Agent tools

What your agent can ask SpringRoll to do.

  1. 01springroll_catalog

    See the applications, releases, deployments, and Data Products you can access.

  2. 02springroll_upsert_application

    Create an application record or update its owner, source, and classification.

  3. 03springroll_upload_bundle

    Send a source bundle when the application does not come from a Git repository.

  4. 04springroll_create_release

    Freeze the current source and provenance as a release.

  5. 05springroll_deploy

    Deploy a release to an environment and return its status and preview URL.

  6. 06springroll_request_approval

    Ask a person to approve production promotion.

  7. 07springroll_request_data_access

    Ask a data owner for a limited, expiring view of a Data Product.

Every tool runs with the organization and permissions you approved in the browser. Production promotion and data access still stop for a person.

Connect company data

Give the app data access without giving it a database password.

An owner publishes a Data Product. The app requests it. A data owner approves the fields, rows, limits, and expiry. SpringRoll's gateway applies those rules to every request; the source credential stays with SpringRoll.

See the controls behind this flow →

Security

The controls stay in the request path.

Security is not a separate review at the end. Tenant checks, human approval, app identity, masking, limits, revocation, and audit evidence run as work moves through SpringRoll.

  1. 01

    Tenant isolation

    Every tenant-owned record carries tenant_id. The scoped data layer injects it, PostgreSQL RLS checks it again, and pages obtain context through verified membership.

  2. 02

    Human approval

    Coding agents can request promotion and data access; they cannot turn model intent into human consent. Production and governed data cross an explicit approval record.

  3. 03

    App Identity

    Each application environment receives a service identity independent from the builder and from the source system.

  4. 04

    Secrets boundary

    Connector and runtime credentials are encrypted at rest and write-only in the interface. Applications never receive source-system credentials.

  5. 05

    Revocation

    Agent connections, access grants, memberships, and application identities can be revoked without changing the upstream source credential.

  6. 06

    Authentication

    Auth.js sessions resolve live memberships on each request. Enterprise identity supports OIDC/SAML and directory provisioning supports SCIM.

  7. 07

    Audit integrity

    An operation and its audit event share one transaction. Events are append-only, ordered per tenant, and hash-chained so removal or alteration is detectable.

These are current product controls. SpringRoll does not claim a certification, compliance status, or uptime figure here.

Start in development

Connect an agent and ship a preview.

Deploy nowSign up