Book a demo

Graph analytics that route themselves, on your infrastructure

Ask a hard question. Get an answer you can audit.

Bridgr reads your files and picks the right analysis for each question. It builds the graph it needs on the fly, then shows you exactly how it got there. No migration. No standing graph database.

Book a demo See how it works
graph route

Five surfaces

  • AskPut the question in plain language.
  • Link AnalysisExplore entities and relationships on the canvas.
  • AnalyzeWatch the notebook build and execute.
  • AnswerRead the finding with its receipt and citations.
  • AutomatePromote a run to a repeatable job. The newest surface.

Every analysis leaves a receipt

Sources read, quality flags, the route the agent chose and why, the scope it worked on, the model revision, and the running cost.

If a graph probe blew its budget and fell back to a table, the receipt says so.

Read 4 files  ·  10,500 work orders  ·  claim rate 6.1%
This question is graph shaped  ·  built a graph  ·  65 nodes
Model revision 3  ·  showing 65 of 2,140
Estimated cost $0.14

What people ask it

  • Service and warranty operations

    Which bays, parts, or techs sit on most of our delayed jobs?

    Bridgr routes it as a graph question, walks the dependencies, and shows the bottleneck with the paths that prove it.

  • Customer and transaction segmentation

    Which customers actually behave alike, and what makes each group different?

    Communities come from behaviour, not from labels someone assigned last year. The notebook that computed them stays open.

  • Document heavy investigations

    Where do the contracts disagree with the invoices?

    Documents are evidence, not a silo. Every claim cites the exact span it came from. Gaps come back labeled inconclusive.

Who builds this

Bridgr is built by Eastridge Analytics, a graph data science consultancy. We are a team of certified Neo4j experts. We find patterns in retail, life sciences, and financial crime data for a living. Founder Tim Eastridge wrote Graph Data Science with Python and Neo4j.

Why it's different

  • Route, do not funnel

    The agent decides whether each question is graph shaped or table shaped. You never pick a tool before you know the answer.

    Route and rationale appear in the receipt on every run.

  • One approval, then it runs

    You confirm the data mapping once. After that the analysis runs start to finish with no mid run checkpoints.

    Mapping approval is the only human gate.

  • Your data lake stays where it is

    Each analysis builds a small, bounded graph from just the slice it needs, then discards it. Nothing is bulk loaded and nothing is migrated.

    Ephemeral per analysis graph, scoped and disposable.

  • Every number traces to a cell

    Answers are synthesized from executed notebook output, not from a model summarizing your files. Open the cell that produced any figure.

    Question battery graded 6 of 6 on grounded in cells.

  • Documents are evidence, not a silo

    PDFs and text are extracted into typed records and graph entities with exact character offsets. Click a claim, land on the source span.

    Citations resolve against the exact source span.

  • Inference is labeled, never disguised

    What the documents actually say is drawn solid and cited. What the system infers beyond the text is drawn dashed and badged. The two never merge.

    Extracted equals cited. Projected equals labeled inference.

Read the FAQ for how it deploys, what it costs, and what it does not claim.