A public ship-log for small, focused experiments in AI devtools, MCP, and developer workflows.
Problem: polished repositories show finished artifacts, but they hide the messy middle. A small experiments repo gives Launchery a place to ship narrow ideas, test assumptions quickly, and turn promising experiments into future standalone tools.
Who this is for: developers exploring MCP, AI tooling, and workflow automation who want compact experiments they can understand in one sitting.
This repo is not a dumping ground for random prototypes. It is a curated lab for:
- experiments that can be explained quickly;
- experiments that strengthen the AI tool builder image;
- experiments that may later become standalone repos, utilities, or examples.
- small developer-facing experiments;
- protocol adapters and bridges;
- automation ideas worth testing before full productization;
- workflow experiments with a clear learning value.
- generic code scraps;
- half-broken playgrounds with no README;
- broad platforms disguised as experiments;
- ideas that do not connect back to the Launchery portfolio.
A narrow experiment showing how a GraphQL API can be exposed through MCP-style tools.
Why it matters:
- demonstrates adapter thinking;
- shows practical MCP extensibility;
- creates a compact demo that is easy to explain.
Explore whether simple workflow definitions can be generated from a narrow input schema.
Experiment with lightweight discovery or inspection of MCP servers/resources.
A lightweight CLI that probes MCP servers for common security issues: input validation gaps, error leakage (paths, stacks, credentials), resource scope problems, and prompt safety. Severity-ranked findings with remediation advice.
experiments/
mcp-graphql-bridge/
README.md
workflow-auto-generation/
README.md
src/generate.ts ← core generator
package.json
tsconfig.json
samples/ ← input specs + expected outputs
mcp-discovery-tool/
README.md
README.md
ROADMAP.md
EXPERIMENTS-PLAN.md
- each experiment has one clear job-to-be-done;
- each experiment is small enough to explain in under 60 seconds;
- every experiment has a README with scope, non-goals, and demo path;
- the repo feels like a coherent lab, not a junk drawer.
- gives Launchery a safe place for lower-risk experimentation;
- feeds future ideas into starter/tools/examples repos;
- signals builder momentum without pretending every idea is a full product.
Blueprint prepared on 2026-04-05. Experiment 2 (workflow auto-generation) implemented on 2026-04-06 — core generator + 3 samples. Experiment 3 blueprint added on 2026-04-06. Next step: push updates to GitHub.