SparkLaunch

SparkLaunch for AI agents

SparkLaunch helps founders validate ideas, build a brand, launch landing pages, manage customer relationships, prepare fundraising materials, and organize company operations through a connected AI assistant.

This repository provides twelve skills and native MCP package metadata for ChatGPT/Codex, Claude Code, Cursor, Gemini CLI, and Muse Code. One canonical source tree generates every host package and native catalog.

Workflow recipes · Contributing · Release status · Security

Get started

Choose your host and follow its setup guide. Protected tools use a SparkLaunch account and host-managed OAuth. Available actions depend on the selected project, your role, granted permissions, and plan or purchased entitlement.

Host	Package	Setup and compatibility
ChatGPT / Codex	OpenAI package	Connection guide and release runbook
Claude Code	Claude package	Local setup and installation
Cursor	Cursor package	Local setup and distribution
Gemini CLI	Gemini package	Setup and account compatibility
Muse Code	Muse skills	Skill installation; protected MCP is disabled in this adapter

Enabled packages use the same Streamable HTTP endpoint:

https://sparklaun.ch/api/mcp/

The packages contain no credentials. Use your host's sign-in flow; do not paste bearer tokens, API keys, client secrets, or authorization headers into skills or MCP configuration. See host compatibility and portable inputs for attachment handling and OAuth evidence boundaries.

What you can do

Skill	Workflows
Platform	Route a founder journey across tools, including monthly reporting and operations.
Projects	Find, select, and manage accessible business projects.
Idea validation	Explore markets, competitors, and TAM/SAM/SOM.
Color palettes	Generate and retrieve brand palettes.
Logos	Generate logos and hand off the files.
Campaigns	Create campaigns, short links, QR codes, attribution, and statistics.
Landing pages	Create and publish pages, then inspect analytics and leads.
Sales CRM	Manage leads, contacts, deals, activities, and supported business-card imports.
SparkCap	Model ownership, dilution, raises, and hiring plans.
SparkRoom	Prepare investor rooms, select document revisions, control sharing, and review usage.
SparkClose	Model SAFEs and manage investment evidence, closing workflows, and signing handoffs.
Incorporation	Prepare eligible cases, private participant tasks, corrections, status checks, and internal Filing Operations submissions.

The recipe catalog connects these skills into complete workflows. Capabilities listed here describe the local candidate; availability in a hosted environment depends on its deployed version and permissions.

Release status

The working-tree candidate is 0.10.0, with a service 1.9.0 snapshot of 109 tools, 33 OAuth scopes, and 12 skills. Its monthly operating cycle adds seven reporting and operations tools and a platform recipe: the host authors investor and board reports while SparkLaunch persists source snapshots, exact approvals, private PDF publication, reviewed room refreshes, and obligations. Existing grants require new consent, and the backend requires additive Founder Close/Operations migrations.

Deployment, hosted package consumption, native-host acceptance, marketplace publication, and real-user outcomes remain unverified for this candidate. The retained service 1.4.0 / 61-tool production observations are historical evidence. See release state, local verification, and earlier candidate notes for the details.

Develop and validate

Use Python 3.12. A standalone SparkLaunch-Skills clone can validate packages using the checked-in tool snapshot:

python -m pip install -r requirements-dev.txt
python scripts/sync_plugin.py
python scripts/generate_submission.py --check
python scripts/validate_host_packages.py
python scripts/validate_submission.py
python scripts/build_submission_bundle.py
python scripts/build_release_bundles.py
python scripts/validate_portal_prerequisites.py --allow-pending
python scripts/validate_public_release.py --allow-pending
python -m pytest -q

See CONTRIBUTING.md for environment setup and the complete source-editing workflow. Edit src/ or adapters/, then regenerate with python scripts/sync_plugin.py --write. Keep generated packages under plugins/ checked in; each host needs its self-contained files.

Clone the application and skills repositories as sibling directories when changing the runtime contract. The exporter defaults to ../SparkLaunch/backend and needs that repository's development dependencies; it uses a temporary local database. The public standalone checks above do not require the application repository.

Installed packages may be cached by plugin version. Published package-content changes require a new governed version; rebuilding locally does not publish or invalidate an existing cache. --allow-pending verifies source consistency while preserving external gates. Follow the release runbook for strict readiness checks, native acceptance, and publishing.

Repository map

Path	Purpose
src/skills/ and src/recipes/	Canonical skill instructions and founder workflows.
adapters/	Host manifests, connection guides, assets, and distribution notes.
contracts/tools.snapshot.json	Runtime-derived tool descriptors and input/output schemas.
plugins/	Generated, self-contained packages for each host.
scripts/ and tests/	Deterministic builds, validators, and regression tests.
evals/	Trigger cases, controlled workflow reviews, and archived evaluations.
docs/	Compatibility notes, release history, and design/verification records.
submission/	Reviewer guidance and recorded release evidence.
release-state.json and server.json	Candidate/distribution status and the MCP Registry descriptor.

Shared safety contract

1. Authentication is managed by the host. Skills never collect credentials or transport headers.
2. projects.list is the source of truth for accessible projects, and project-scoped tools receive an explicit project_id.
3. Every write uses one stable idempotency_key for the exact mutation. An uncertain write is read back before any retry.
4. Project plan, project role, OAuth scope, and commercial entitlement are separate gates.
5. Destructive or public-state tools expose truthful annotations and use the runtime's governed preview/confirmation or version/idempotency boundary.
6. Generated files use short-lived HTTPS references, never raw base64 or data URLs.
7. Identifiers and concurrency versions remain opaque tool-call state; user-facing output uses human-readable names.
8. Configured assets, published state, traffic, conversion, CRM persistence, formation status, deployment, and marketplace publication are separate proof layers.

Support and license

• Support: support@sparklaun.ch
• Security reports: SECURITY.md
• Privacy: SparkLaunch Privacy Policy
• Terms: SparkLaunch Terms and Conditions
• License: Apache License 2.0, with NOTICE

This repository is licensed under Apache-2.0. The separately hosted SparkLaunch service and backend remain proprietary; service terms, account permissions, and trademark rights are unchanged.