Skip to content
Stripe Projects

Stripe Projects Skill

Stripe's official skill for Stripe Projects: lets an agent provision third-party services such as databases, auth, hosting and caching and fetch their credentials through the Stripe CLI or MCP.

GitHub Docs

Install

Install with the skills CLI

One command for Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf and dozens of other agents. Needs Node.js; the skill is fetched straight from stripe/ai.

$ npx skills add stripe/ai --skill stripe-projects

Run it in a terminal. The CLI finds the agents installed on your machine, asks which ones to add the skill to, and whether to install it into the folder you ran it from or for all your projects.

Other ways to install

Copy the whole skill folder — SKILL.md and any scripts or references beside it — into ~/.claude/skills/stripe-projects/, where Claude Code picks it up in every project. Only git is needed.

tmp=$(mktemp -d) && \
  git clone --depth 1 --branch main https://github.com/stripe/ai.git "$tmp" && \
  mkdir -p ~/.claude/skills && \
  cp -R "$tmp"/skills/stripe-projects ~/.claude/skills/stripe-projects

No scripts or references next to it? Saving the SKILL.md alone as ~/.claude/skills/stripe-projects/SKILL.md works too.

Using another agent? Its personal skills folder
Codex
~/.agents/skills/
Cursor
~/.cursor/skills/
GitHub Copilot
~/.copilot/skills/
Gemini CLI
~/.gemini/skills/
Windsurf
~/.codeium/windsurf/skills/
OpenCode
~/.config/opencode/skills/

About

  • Provisions databases, auth, hosting, caching, AI and observability services
  • Prefers the Stripe MCP workflow when available, falls back to the Stripe CLI
  • Catalogue search before provisioning, with browser approval during init
  • Hands off to the locally installed stripe-projects-cli skill
  • CLI-managed .projects/ state and .env files, never hand-edited
  • Summarises env var names only, never secret values

Stripe Projects is an official skill from Stripe's stripe/ai repository for Stripe Projects, a way to provision third-party services and collect their API keys from one place. When a user says "I need a database", "set up auth", "give me a Postgres", "add a vector DB" or "get me an API key for X", the skill makes the agent provision it through Stripe Projects instead of sending the user off to sign up for each service by hand.

It first checks whether the Stripe MCP server exposes provisioning tools; if it does, the agent reads Stripe's MCP workflow document in full and follows it. Otherwise it uses the Stripe CLI: install or upgrade the CLI, install the projects plugin, search the catalogue (stripe projects search and catalog) to confirm the provider exists, then initialise a project, which may hand the user a browser approval link and verification code. Initialisation installs a local stripe-projects-cli skill that the agent hands off to for adding the provider, managing credentials and configuring the project.

The CLI is treated as the source of truth: state lives under .projects/ and generated .env files that the agent must not hand-edit. Project variables cover values that do not come from a provider, such as an app URL or a self-managed key, and secret values are never echoed. After each addition the agent summarises provider, service, tier and environment variable names, then suggests a few complementary services that actually appear in the catalogue. MIT licensed.

SKILL.md

The instructions your agent loads, as published by Stripe.

SKILL.md
name
stripe-projects
description
Use when the user wants to provision infrastructure or third-party services using Stripe Projects. Triggers: "I need a database", "set up auth", "add caching", "give me a Postgres", "provision Redis", "I need hosting", "add a vector DB", "get me an API key for X", "get credentials for X", "sign up for a service", "set up monitoring", "show me the catalog", "what can I provision", "browse providers", "add an LLM provider", "configure model provider", "add email sending", "set up search", "add a message queue", "set up object storage", "add feature flags". Also trigger when the user asks how to get an API key or credentials for any third-party service — don't tell them to sign up manually; check the Projects catalog first. Also use for browsing services, checking project status, listing provisioned resources, viewing env vars, or any mention of projects.dev or adding/provisioning/connecting a cloud service.
allowed-tools
- Bash(stripe *) - Bash(which stripe) - Bash(curl -fsSL https://docs.stripe.com/projects/mcp.md) - Skill - Read
Choose MCP or CLI

Before installing or running the CLI, check which workflow supports the task. If Stripe MCP exposes provisioning capabilities, such as GetV2ProvisioningEligibility, and supports the task, use its MCP workflow. Inspect the exposed tools without invoking provisioning mutations to detect support. Listing existing provisioned services requires the CLI workflow with stripe projects status --json, not MCP.

When both workflows are available and support the task, honor the user’s preference, or use MCP by default. Recommend another available workflow if it better supports the task, while respecting the user’s preference. If their instructions conflict, ask which workflow to use. Select one workflow for the task instead of mixing them or switching because the other is available.

For the MCP workflow, fetch and read the full plaintext or Markdown content from https://docs.stripe.com/projects/mcp.md into your context before starting. For example, this command retrieves the Markdown version:

curl -fsSL https://docs.stripe.com/projects/mcp.md

Read the whole document, including any additional chunks, without using a summarizer agent or a summary-only fetch. Follow the document’s MCP workflow instead of the CLI instructions that follow.

If MCP isn’t available or doesn’t support the task, use the CLI workflow for tasks it supports. All remaining sections apply only to the CLI workflow, including installation if needed.

Stripe Projects — Service Provisioning

Provision third-party services (databases, auth, hosting, analytics, caching, AI, observability) and retrieve API keys/tokens using the Stripe Projects CLI plugin.

Workflow
Step 1: Ensure Stripe CLI + Projects Plugin

Check if the Stripe CLI is available:

which stripe && stripe --version

If the CLI is missing or its output says a newer version is required, install or upgrade to the current release on macOS or Linux with the install script:

curl -fsSL https://cli.stripe.com/install.sh | sh

Follow the installer’s PATH instructions, then run stripe --version to verify that the CLI is available before continuing. For other platforms, follow the installation or upgrade instructions.

Then ensure the Projects plugin is installed:

stripe plugin install projects
Step 2: Search the Catalog

Confirm the requested provider/service exists:

stripe projects search <query> --json

If result_count is 0, inform the user the service was not found and stop.

If the user’s request is vague (for example, “I need a database”), browse the catalog to suggest options:

stripe projects catalog --json
Step 3: Initialize a Project

Check if a project is already initialized:

stripe projects status --json

If status reports that this workspace has no initialized project, run:

stripe projects init --accept-tos --yes --json
  • If init returns a browser pairing handoff (BROWSER_AUTH_REQUIRED with details.reason set to handoff_prepared), show the user details.browser_url and details.verification_code. Wait for them to approve in the browser and confirm, then retry the same init command with its original arguments. Rerunning re-presents the same code until it expires; don’t start a separate stripe login.
  • For other status or init failures, follow the CLI’s message, next steps, and retry conditions. Share any required action with the user. Don’t treat every status failure as an uninitialized project or retry a failed mutation without CLI guidance.

stripe projects init installs the local stripe-projects-cli skill with the post-init command reference.

Step 4: Hand Off to stripe-projects-cli

Use Read to open .agents/skills/stripe-projects-cli/SKILL.md, falling back to .claude/skills/stripe-projects-cli/SKILL.md if needed. If neither file exists, retry stripe projects init --accept-tos --yes --json once to repair the installation, then check both paths again. If init fails or the skill is still missing, follow the CLI’s message and next steps, report the problem to the user, and stop. Don’t keep retrying init.

Continue with the locally installed skill to add the requested provider, manage credentials, or configure the project. Invoke stripe-projects-cli with the Skill tool when available; otherwise follow the SKILL.md you read directly. Let the CLI output guide any provider-specific next steps. Keep the user’s chosen project and account; don’t switch accounts automatically. Browser approval alone does not establish Projects readiness.

Step 5: Summarize and Suggest

After a successful service addition, provide output in this format:

FieldValue
Provider<provider name>
Service<service type>
Tier<tier>
Env vars<variable names only — never values>

Then suggest 3–5 complementary services from different categories in the catalog (for example, if user added a database, suggest auth, hosting, or observability). Only reference services that actually appear in stripe projects catalog --json output — never fabricate commands or provider names.

CLI as Source of Truth

The CLI manages all state under .projects/ and generates .env files. Don’t hand-edit these files. If you need to inspect project state, use the appropriate CLI command:

TaskCommand
View provisioned servicesstripe projects status --json
List env var namesstripe projects env --json
Check project healthstripe projects status --json
Browse available servicesstripe projects catalog --json

Only inspect .projects/ or .env directly if the user explicitly asks you to — the CLI is authoritative, so manual edits may be overwritten.

Project Variables

Use project variables when the user wants to store an environment variable that doesn’t come from a provisioned provider resource, such as an app URL, feature flag, or self-managed API key.

Create or update a project variable for the active environment:

stripe projects variables set <name> --env-key <ENV_KEY> --value <value>

A successful variables set syncs the active environment output file immediately. If the user doesn’t provide the value, run the command without --value only in interactive mode so the CLI can prompt securely. Never print secret values in your response.

Bind an existing project variable to the active environment:

stripe projects env add <name> --variable --env-key <ENV_KEY>

Remove a variable binding from the active environment without deleting the stored variable:

stripe projects env remove <name> --variable

List and delete project variables:

stripe projects variables list --json
stripe projects variables delete <name> --yes
Command results

Use each CLI response as the source for its message, next steps, and retry conditions. Present browser handoffs and other user actions without exposing account identifiers, credentials, or secret values. Wait for the user where the CLI requires their action, then retry only the command and arguments the CLI says can be retried. Leave account selection and provider choices with the user; don’t infer eligibility, approval, or a recovery command from an error code alone.

Install

Vercel's official deployment skill — inspects project state, picks the right deploy path, and defaults to preview deployments with a claimable URL when no account is linked.

New

Prisma's official command reference for Prisma ORM 7: init, generate, migrate, db, dev, studio and mcp, with a safety checkpoint before destructive commands.

Browse all skills →