SparkLaunch

SparkLaunch helps early-stage founders turn ideas into incorporated, growing, and investor-ready companies. Through guided workflows, AI-assisted tools, founder programming, and continuously updated readiness metrics, founders can validate customer needs, plan their product and go-to-market strategy, incorporate, manage equity and company documents, and prepare for fundraising.

GitHubTry this MCPSponsored

Documentation

SparkLaunch Skills

This repository contains the canonical skills, recipes, and installable plugin package for the connected SparkLaunch ChatGPT experience.

User Experience

SparkLaunch helps founders select or create a business project, validate an idea, generate brand assets, publish measurable launch surfaces, review campaign and landing-page signals, operate private CRM workflows, and prepare entitlement-gated incorporation cases with person-specific Action Center tasks.

The default broad workflow is:

  1. Connect SparkLaunch and select an accessible project, or create one with a complete business description.
  2. When a project is created, wait for its automatically queued Idea Validation research to complete (normally 10-15 minutes).
  3. Generate palette and logo options.
  4. Create a campaign, QR file, and landing page.
  5. Review observed signals and grounded CRM context.

Install And Connect

  • ChatGPT: SparkLaunch availability in ChatGPT is separate from this public repository and the MCP Registry listing. When the SparkLaunch app is available to your account, start a new conversation with SparkLaunch selected. OAuth begins on the first protected action.
  • Codex: Add https://github.com/SparkLaunch-Dev/SparkLaunch-Skills as a Git plugin marketplace and install the sparklaunch plugin. Open a new task after installation so the current plugin actions are loaded.
  • Other MCP clients: Discover io.github.SparkLaunch-Dev/sparklaunch through the official MCP Registry or configure the Streamable HTTP endpoint https://sparklaun.ch/api/mcp/. The client must support the server's OAuth flow; never paste tokens into prompts or configuration shared with other people.

See Connect SparkLaunch to ChatGPT for connection, reconnection, project selection, and disconnection guidance.

Shared Safety Contract

  1. Authentication is OAuth-connected and managed by the host. It starts on the first protected SparkLaunch action in a conversation where the connector is loaded; skills never request credentials, OAuth codes, or transport headers. If required actions are absent, the skill stops and directs the user to open a new ChatGPT conversation with SparkLaunch selected instead of misreporting connector absence as an OAuth challenge. If a loaded connection is expired or revoked, the skill stops before writes, asks the user to reconnect from the AI Agent, and retries only after reconnection succeeds and any uncertain prior result is checked.
  2. projects.list is the source of truth for accessible projects. Project-scoped tools receive an explicit project_id argument.
  3. Every write receives one stable idempotency_key for the exact intended mutation. Uncertain writes are not repeated with new keys.
  4. Before a write or confirmation, the skill checks projects.get.effective_permissions when project-scoped authorization applies. Destructive or public-state tools return a one-time confirmation preview only after server-side authorization preflight succeeds. The user must explicitly approve it before the exact call is resubmitted with its confirmation token.
  5. Skills do not use query-token REST URLs, compatibility endpoints, legacy project headers, or hidden fallback routes.
  6. Generated logo and QR outputs use short-lived HTTPS file references, never raw base64 or data URLs.
  7. Configured assets, published state, traffic, conversions, CRM persistence, and revenue are reported as separate proof layers.
  8. User-facing errors stay concise. Secrets, private diagnostics, internal ownership IDs, and unnecessary personal data are never surfaced.
  9. OAuth scopes are the connection's maximum authorization. The selected project's plan and the user's project role may further restrict a tool; a plan/role denial is not a reason to reconnect OAuth.
  10. Identifiers and concurrency versions remain available only as opaque internal tool-call state. User-facing prose, tables, confirmations, and handoffs use human-readable names or descriptions and never expose identifier/version values, labels, parenthetical references, or columns.

Repository Layout

  • sparklaunch-*/: canonical skill source
  • recipes/: multi-tool connected founder workflows
  • plugins/sparklaunch/: deterministic packaged mirror plus plugin metadata
  • .agents/plugins/marketplace.json: repository-local install catalog

The canonical skill folders are the only files edited by hand. Run scripts/sync_plugin.py --write to update packaged mirrors and scripts/validate_submission.py to check encodings, parity, metadata, tool references, and submission artifacts.

Development And Validation

Clone the application and skills repositories as sibling directories when running the full package tests or regenerating the ChatGPT submission:

parent/
  SparkLaunch/
    backend/
  SparkLaunch-Skills/

scripts/generate_submission.py imports the canonical MCP tool contracts from ../SparkLaunch/backend, so python scripts/generate_submission.py --check and python -m pytest tests -q require that sibling checkout. A standalone SparkLaunch-Skills clone can still run python scripts/validate_submission.py for package-only validation.

Every published plugin change must also update plugins/sparklaunch/.codex-plugin/plugin.json to a new version. Git marketplace installs are cached by plugin version, so publishing changed files under an existing version can leave an older cached package active.

Official MCP Registry

The root server.json publishes the production service as the remote Streamable HTTP server io.github.SparkLaunch-Dev/sparklaunch. The descriptor version must match ../SparkLaunch/backend/mcp_server_version.py; bump both for every new Registry publication because published versions are immutable.

The descriptor intentionally omits repository: the public skills repository is not the private MCP server source, and the official Registry supports a closed-source server when its remote endpoint is publicly accessible. Validate from this repository with the latest official publisher:

mcp-publisher validate

Publish an approved new version with the repository's Publish to MCP Registry GitHub Actions workflow. It downloads the pinned publisher release, verifies its checksum, validates server.json, authenticates with GitHub OIDC, and publishes without a stored Registry credential. Registry versions are immutable, so deploy and observe the matching application-owned version before dispatching the workflow.

Version 1.0.0 is currently published and active. Verify discovery through:

https://registry.modelcontextprotocol.io/v0.1/servers?search=io.github.SparkLaunch-Dev/sparklaunch

MCP Registry publication, GitHub availability, production deployment, and ChatGPT app review are separate release states. Do not describe one as proof of another.

Current Skills

  • sparklaunch-platform: broad founder-workflow router
  • sparklaunch-projects: project discovery and management
  • sparklaunch-idea-validation: market, competitor, and TAM/SAM/SOM analysis
  • sparklaunch-color-palettes: palette generation and retrieval
  • sparklaunch-logo-generation: logo generation and file handoff
  • sparklaunch-campaigns: campaigns, short links, QR, attribution, and statistics
  • sparklaunch-landing-pages: landing creation, publishing, analytics, and leads
  • sparklaunch-sales-crm: lead, contact, deal, activity, and business-card workflows
  • sparklaunch-incorporation: entitlement, case preparation, private participant tasks, corrections, status, and internal Filing Operations submission

See recipes/README.md for the supported multi-step workflows.

Support, Security, And License