> ## Documentation Index > Fetch the complete documentation index at: https://metalworks.lab2a.ai/llms.txt > Use this file to discover all available pages before exploring further. # metalworks > Go from a startup idea to launch, grounded in what people actually want. Validate demand, get positioning, a design system, a build plan, and launch copy — every claim backed by a real quote you can open. **metalworks turns a startup idea into a launch plan, grounded in real demand.** Give it one sentence about what you want to build. It reads real conversations — Reddit, Hacker News, the web, or your own data — to tell you whether people actually want it, then turns that into the things you need to launch: your positioning, the competitors to beat, a design system, a build plan for your coding agent, and launch copy. **Every claim links back to a real quote you can click — nothing is invented.** It's a Python library you install (also a CLI, an MCP server, and a Claude Code plugin). ## Your first report Install metalworks with a provider SDK, and set **one** key — any provider works (embeddings default to a local model, so there's no second key to set): ```bash theme={null} pip install "metalworks[research,openai]" # or [research,google], [research,anthropic] export OPENAI_API_KEY=... # or ANTHROPIC_API_KEY / GOOGLE_API_KEY ``` Or **run keyless on your Claude Code login** — no key at all (the chat model and web search use your Claude Code session): ```bash theme={null} pip install "metalworks[research,claude-code]" # then just run; nothing to export ``` ```text Claude Code theme={null} /demand-report an affordable, jitter-free focus supplement for developers ``` ```python Python theme={null} from metalworks import Metalworks mw = Metalworks() # provider inferred from your env key research = mw.research( "an affordable, jitter-free focus supplement for developers", subreddits=["Nootropics", "Supplements"], ) report = research.demand # .research() returns a report print(report.demand_summary) # how strong is the demand? a one-line summary for cluster in report.ranked_clusters: print(cluster.distinct_author_count, "people:", cluster.claim) for quote in cluster.quotes: # the real comments behind it — nothing invented print(" ", quote.source_url) ``` That's a real report grounded in real conversations you can open — see the full [Quickstart](/docs/quickstart). ## What you can do with it Run demand research once, then turn that one report into whatever you need next. The work falls into five stages — each step is a single call: | Stage | What you get | | -------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Research** | [Demand research](/docs/demand-research) — a read on how strong the demand is + the real needs people voiced, each backed by quotes. Run the full [validation loop](/docs/validation-loop) — frame an idea, weigh demand against what already exists, get a **GO / PIVOT / NO-GO** call — then your [positioning](/docs/positioning) (who it's for, why it's different) and the [competitors](/docs/competitors) to beat. | | **Design** | [A design system](/docs/design-system) — an aesthetic direction and a SAFE/RISK choice per dimension, read from what rivals actually do. | | **Build** | [A build spec](/docs/build-spec) — what to build (surface + screens), a feature plan mapped to real demand, and a project scaffold your coding agent executes. | | **Distribute** | [Distribution](/docs/distribution) — **test→focus** channel experiments, channel-shaped drafts, and a corpus-derived data report, each line backed by a quote (it never posts) — plus [GEO / LLM-citability](/docs/distribution-geo): the real threads to join and answer-first briefs so people and AI search cite you. | | **Grow** | [Reddit engagement](/docs/reddit-engagement) — find threads to join and draft honest replies (you approve every post). | New here? Walk one idea from start to finish in the [end-to-end walkthrough](/docs/walkthrough). ## Why it's not just another AI tool Most AI research tools make up plausible-sounding answers. metalworks doesn't. It reads real conversations across real sources, and **anything it can't back with an actual quote, it drops.** When the demand isn't there, it tells you — instead of inventing an opportunity. That's what makes the output safe to act on. See [why you can trust the output](/docs/how-it-works). ## Four ways to use it The same engine, whichever fits your workflow: * **Python library** — `from metalworks import Metalworks`. The full toolkit, embeddable. → [SDK reference](/docs/python-sdk) * **CLI** — `metalworks research run --question "..."`, then `metalworks build init`, etc. → [CLI](/docs/cli) * **MCP server** — `metalworks mcp serve`. Drive it from any MCP-aware agent or app. → [MCP tools](/docs/mcp-tools) * **Claude Code plugin** — `/demand-report`, `/build-spec`, and more as slash commands. → [Claude Code plugin](/docs/claude-code) The docs are split to match: **Documentation** (this section, the workflow end to end), **Claude Code & plugins** (the slash commands), and the **SDK reference** (the Python API and how to extend it). ## Install ```bash theme={null} pip install "metalworks[research]" # the demand-research pipeline ``` Need only the data tools, or want to keep it lean? See [Installation](/docs/installation) for the full options. metalworks is open source (MIT) on [GitHub](https://github.com/Lab2A/metalworks).