A spike is a short, time-boxed piece of research done to answer one technical question before you commit to it: which state library, which ORM, which testing setup. SpikeMe generates that document for you, grounded in your actual stack and in live registry data, so the decision is fast, opinionated, and backed by real numbers.
How it works
You give SpikeMe your project manifest (package.json for npm, pyproject.toml
or requirements.txt for Python, go.mod for Go). It:
- Analyzes your stack locally: parses the manifest, categorizes your dependencies, and spots gaps that make good spike topics. Nothing is uploaded for this step.
- Generates a spike document: you pick a topic and depth, and SpikeMe writes a full decision doc: options analyzed, a comparison table, a clear recommendation, a proof-of-concept plan, and risks.
- Grounds it in evidence: before writing, it collects live facts (current version, weekly downloads, license, bundle size) from the npm registry, PyPI, or deps.dev and injects them, so the document cites real, dated numbers.
Three ways to use it
- Web: the wizard at spikeme.io/app: upload a manifest, pick a topic, read and share the document.
- CLI:
npx spikeme: analyze your repo and generate spikes from the terminal, where your code already lives. - MCP: the SpikeMe MCP server exposes
generate_spiketo your coding agent (Claude Code, Cursor), so the agent produces the artifact it otherwise can't.
Open core
The analysis is open and local: parsing, categorization, and gap detection run on your machine and are free. Generation runs through the SpikeMe backend (it uses our models and the evidence engine) and is metered by your plan.
Next steps
- Installation: get the CLI or open the web app.
- Generating spikes: depth, options, and the evidence engine.
- Ecosystems: what SpikeMe understands today.