How Signals works.
A plain guide to Workspaces, Sessions, signals, sources, and what your team can do at each step.
What Signals is
A signal is a candidate indicator of change, something worth tracking because it might tell you where a market, technology, or behavior is heading. Signals helps research teams discover those indicators, source them against real-world evidence, score them against the dimensions that matter, and synthesize them into reports.
The platform combines a structured research workspace with AI agents that do the heavy lifting: searching the web, deduping findings, judging supporting and contradicting sources, and writing up the results. You stay in the loop the whole time: you can edit signals by hand, chat with an analyst agent, and curate findings across sessions.
Core concepts
Each concept has its own page. Open it to unfold the details, gotchas, and how it relates to everything else.
- Organization
- The top-level account that owns every workspace, session, and signal.
- Workspace
- A container for related research that defines the structure every session inside it uses.
- Playbook
- A reusable research blueprint of categories, metrics, and workflow, deep-copied into a workspace.
- Session
- A single research engagement inside a workspace, with its own briefing, signals, and reports.
- Briefing
- The session’s objectives, scope, and constraints: the prompt that steers the research.
- Workflow
- An ordered sequence of pipeline tasks defined on the workspace and triggered per session.
- Signal pipeline
- The six-stage AI pipeline that produces signals and reports from a briefing.
- Frame
- A reusable lens applied to signals: a way of grouping or projecting findings without changing them.
- Tag
- A free-form label attached to a signal, drawn from a workspace-defined tag set.
- Metric
- A scoring dimension applied comparatively to every signal in a session.
- Signal
- A candidate indicator of change: the structured atomic finding of Signals.
- Source
- A URL with supporting context attached to a signal, judged for or against the claim.
- Grounding
- The four-state lifecycle that tracks how well-supported each signal is.
A session, end to end
- Create a workspace from a playbook. This sets up the categories, metrics, report types, and workflows your research will use.
- Start a session inside the workspace. Give it a title and write a briefing: what are you investigating, for whom, with what constraints.
- Run a workflow. Pick one of the workspace’s workflows and choose auto (runs every task end-to-end) or manual (you click through, task by task).
- Review the signals as they accumulate. Edit titles, summaries, categories, tags, and sources by hand whenever you want.
- Read the report when the pipeline writes one. Re-run the workflow whenever you want to refresh.
- Curate across sessions by pulling standout signals into collections.
The signal pipeline
A typical workflow chains six stages. Each is a task you can configure on the workspace, re-order, or omit. Open the pipeline page for the per-stage detail.
Search the web and synthesize candidate signals against the briefing.
Consolidate duplicates; keep originals soft-merged for audit.
Re-check sources, judge each, commit to a signal-level verdict.
Score every active signal comparatively against each metric.
Render a per-signal image from a prompt template.
Synthesize all active signals into a markdown narrative.
Tasks skip work that’s already complete by default. Re-running a workflow only redoes what changed.
Ready to start? Sign in to your organization.