> ## Documentation Index > Fetch the complete documentation index at: https://atlas.rebuildingus.org/docs/llms.txt > Use this file to discover all available pages before exploring further. # How Atlas works > How public sources become searchable, source-linked civic profiles. Atlas turns public evidence into a directory people can search, inspect, and act on. The workflow is designed around one user outcome: a person should be able to understand who is doing civic work in a place and why Atlas believes the record is worth showing. ## The short version Atlas begins with a question such as "Who is working on housing in Detroit?" or "Which organizations are doing transit advocacy in Nevada?" Atlas searches local journalism, organization websites, nonprofit directories, public records, campaign pages, reports, and other public material. The discovery engine reads sources and looks for named people, organizations, initiatives, campaigns, and events tied to the place and issue. Each candidate record keeps the source title, publication, URL, date, source type, and the source context that explains why the actor was extracted. The same actor may appear in several sources. Atlas merges likely duplicates and uses source density, recency, specificity, and contact surface to decide how strongly to present the record. Atlas should make thin, stale, partial, and uncorroborated areas visible. A missing or weak result is a research signal, not something to hide. ## What counts as evidence Atlas can use different kinds of public evidence: | Source type | Example | What it can support | | -------------------- | ------------------------------------------------------- | ----------------------------------------------------------------------------- | | Local journalism | A feature on tenant organizing in Detroit. | Names, roles, issue areas, places, and current activity. | | Organization website | A program page from a worker center. | Mission, services, geography, public contact surface, and active initiatives. | | Public record | A city agenda, grant list, or public meeting record. | Participation, official status, dates, and jurisdiction. | | Directory or roster | A coalition list, partner page, or nonprofit directory. | Existence, affiliation, geography, and contact surface. | | Research or report | A civic landscape report or university paper. | Context, issue framing, named actors, and broader ecosystem signals. | A source does not have to be a long article. A roster row, partner list, contact block, table, or public record can be useful evidence when the context is clear enough to show why the actor belongs in Atlas. ## How a profile becomes useful A useful profile is not just a name. It should answer four questions: 1. **Who is this?** The profile should identify a person, organization, initiative, campaign, or event. 2. **Where is the work happening?** The profile should connect the actor to a place at the right level of specificity. 3. **What issue does it touch?** The profile should connect to one or more issue areas without pretending the source proves more than it says. 4. **Why should I trust this?** The profile should expose the source evidence, dates, and confidence signals behind the record. Atlas people directory showing profile cards with locations, issue tags, source counts, and updated dates. ## Why discovery runs report gaps Atlas is useful when it shows what it found and what it did not find. A team looking at Phoenix may have strong evidence for worker centers and immigrant rights groups, but weak evidence for environmental justice or broadband access. That gap matters because it tells the team where manual research, a new Scout run, or a local partner conversation should happen next. Gap reporting also protects the user experience. Atlas should not imply that a place has no civic work on an issue just because one search pass did not find enough public evidence. ## Ways to run discovery | Method | Best for | Where to start | | ------------------- | -------------------------------------------------------------------- | --------------------------------------------- | | Public search | Looking up known public records and profiles. | [Browse Atlas](/docs/quickstart-browse) | | Workspace discovery | Team research runs, saved actors, briefs, and coverage review. | [Team workflow](/docs/workspace/customer-workflow) | | Scout | Local or scripted research before syncing selected results to Atlas. | [Scout CLI](/docs/scout/overview) | | MCP | Conversational research through an AI assistant. | [MCP overview](/docs/mcp/overview) | | REST API | Programmatic search, source review, and workflow automation. | [API guide](/docs/api) | Atlas discovery screen with a form for location, state, issue areas, and a recent-runs panel. ## The trust boundary Atlas organizes public evidence; it does not certify that every source is complete, current, or correct. That is why source links, dates, extraction context, correction paths, and confidence signals are part of the core product experience. Read [Trust and provenance](/docs/resources/trust) before using Atlas for decisions that affect real people, outreach, publication, funding, or public claims. ## What's next What Atlas knows, how it knows it, and how to interpret uncertainty. Find housing organizers in Detroit and inspect the source evidence. Build a source-linked briefing for a nonprofit entering Phoenix. Run Scout locally, sync to Atlas, and open the workspace run.