BetaEditorial Listing

Waldo by Wherobots

Wherobots · Waldo by Wherobots: Geospatial AI Agent for Spatial SQL, Satellite Imagery, Maps and Pipelines

Open Waldo by Wherobots

Waldo is the geospatial AI agent built into Wherobots Cloud. You ask a question about places in plain English, and it finds vector and raster datasets in your catalogs, writes and runs spatial SQL, Python and imagery-inference tasks on Wherobots, maps the result and can turn the analysis into a scheduled pipeline. It suits data engineers, data scientists and GIS analysts who work with large spatial datasets or satellite and aerial imagery.

PricingPaid
Setupmedium
Runs onWeb
APINo
Open sourceNo
DocsYes
CategoryProductivity
GeospatialGISSpatial SQLData EngineeringData ScienceSatellite ImageryEarth ObservationData PipelinesApache SedonaIcebergLakehouseMappingPublic Preview

Best for

Data engineers, data scientists and GIS analysts whose spatial data or imagery already lives in cloud storage or a lakehouse, who want an agent that finds the right datasets, writes and runs spatial SQL and imagery inference at regional or national scale, and turns the result into a scheduled pipeline

Not ideal for

Teams that are not willing to run work on Wherobots Cloud, desktop GIS users who mainly edit and style maps by hand, organizations that need prompts processed in a single region, and anyone who wants to call the agent from their own app through an API

Who it's for

Data engineers, data scientists and GIS analysts working with large vector datasets and satellite or aerial imagery in cloud storage and lakehouses

Capabilities

  • Natural-language chat in Wherobots Cloud, available as an Ask Waldo sidebar on every page or a dedicated chat page, that asks up to three clarifying questions with suggested defaults before it starts
  • Data discovery across the Havasu Catalog, Wherobots Open Data such as Overture Maps, your own tables and federated catalogs such as Databricks Unity Catalog, AWS Glue and Amazon S3 Tables, checking schemas before it writes a query
  • Writes and runs spatial SQL and Python on WherobotsDB, Wherobots' raster and vector query engine with 300+ spatial functions, and lists each query so you can check it
  • Runs multi-step workflows that combine WherobotsDB with RasterFlow, Wherobots' imagery inference engine, which supports models such as Meta's SAM 3 for text-prompted object detection
  • Works with vector formats (GeoParquet, GeoJSON, Shapefile) and raster formats (GeoTIFF, COG, Zarr), with integrations for Iceberg, Databricks, Amazon S3, AWS Glue, ESRI and STAC
  • When a dataset is not in your catalogs, it can find where it is published and, after you approve the job, ingest it as a table in your Havasu Catalog; web search for public facts and datasets is on by default where available and can be turned off by an admin
  • Returns answers with a summary, tables, map cards that open in Wherobots Map, and the datasets, definitions and data dates used, with every step and its code viewable under Worked
  • Turns an analysis into a scheduled pipeline: writes a Python script, creates the schedule paused, tests one time window and resumes it after you approve, writing versioned tables to Iceberg-compatible platforms
  • Approval prompts before it starts or stops runtimes, runs jobs, writes or deletes data, creates Python files or changes existing files, saves maps, runs billed RasterFlow tasks or creates or changes schedules, with Approve, Always allow, Deny and Always deny options; jobs show their runtime size and hourly cost range first
  • Every query and tool call runs under the signed-in user's own Wherobots permissions, so Waldo can read only the catalogs, tables and files that user can access

Limitations

  • In public preview since November 18, 2026; Wherobots says features and limits may change before general availability
  • Works only inside Wherobots Cloud on data reachable from its catalogs and storage integrations; it is not a desktop GIS tool
  • AI token quotas apply: 55 million tokens one-time for trial organizations or those without a payment method, and 140 million tokens per week once a payment method is added; reaching the quota stops processing
  • AI tokens are free during the preview, but every WherobotsDB and RasterFlow run it starts is billed at your normal compute rates, including scheduled runs when no one has Waldo open, and per-chat costs are not shown in the chat yet (you check Usage History instead)
  • Cannot execute notebooks or read their cell outputs
  • Needs clear definitions for ambiguous terms such as 'downtown' or 'tower sites'; Wherobots recommends starting with smaller areas, such as a single county, and verifying results
  • Schedules run a Python script or JAR at least 15 minutes apart; RasterFlow tasks cannot be scheduled directly, a schedule is auto-disabled after 5 consecutive failed runs, and task schedules may not be enabled for every organization
  • The default model uses an Amazon Bedrock global inference profile, so a prompt can be processed in any AWS commercial region, which may matter for strict data residency requirements
  • RasterFlow currently runs in AWS US West (Oregon), and its pricing may change during or after its own public preview
  • No API or MCP access to Waldo itself is documented, and it cannot read local files or code repositories; the Wherobots MCP server and CLI are for connecting external agents such as Claude Code or Cursor instead

Use cases

  • Counting buildings and power assets inside historical hurricane or flood footprints and mapping the exposure
  • Detecting objects such as solar panels or rooftops in aerial imagery for a city and joining the results to parcel data
  • Computing a vegetation index from Sentinel-2 imagery for every field in a state and mapping the weakest fields
  • Flagging land change around a portfolio of sites each time new satellite imagery arrives, as a scheduled pipeline
  • Mapping GPS trips that ran off route or late from fleet data and refreshing the analysis on a schedule
  • Writing a cleaned, versioned spatial data product back to an Iceberg lakehouse for other engines to query

Our take

Waldo's case is that spatial work fails less often at writing code than at everything around it: finding the right layers, getting coordinate systems and spatial joins right, running heavy jobs at state or country scale, and turning a one-off answer into something that refreshes. Because it sits inside Wherobots Cloud, it can search the catalogs, check schemas, run spatial SQL and imagery inference on Wherobots' own engines, and schedule the result, with an approval step before anything that costs money or writes data. That makes it far more useful than a general chatbot for teams already storing spatial data in S3, Iceberg, Databricks or Glue. The tradeoffs are clear in Wherobots' own docs. It only works on Wherobots, it is a public preview with token quotas, compute is billed for every run with no per-chat cost total in the chat yet, and prompts can be processed in any AWS commercial region. If your spatial data already sits in a cloud lakehouse, the free trial is a cheap way to test it on one county-sized problem before trusting it with recurring pipelines.

Who should use it

Data engineers and data scientists who need to join, analyze and refresh large spatial datasets in a cloud lakehouse; GIS analysts who want to go from a question to a map without writing spatial SQL by hand; and teams that run imagery analysis on satellite or aerial data, such as insurance, energy, agriculture or logistics teams working on exposure, crop, change-detection or fleet analytics.

Who should skip it

Teams that need their spatial work to stay on desktop GIS or their own infrastructure, organizations that need prompts processed in a single region, users who mainly need cartographic editing rather than analysis, and developers who want to embed an agent in their own application through an API.

Strengths

  • Purpose-built for spatial data: it searches spatial catalogs, checks schemas and uses WherobotsDB's 300+ spatial functions and RasterFlow imagery inference rather than only generating code
  • Handles both vector data and satellite or aerial imagery in one conversation
  • Turns a one-off analysis into a scheduled pipeline that writes versioned tables back to your Iceberg-compatible lakehouse
  • Runs under each user's own data permissions and asks before starting runtimes, running jobs, writing data or running billed tasks, showing the job's hourly cost range first
  • AI tokens are free during the preview, and the Professional trial needs no credit card
  • Wherobots says it does not train models on your prompts, chat history, query results or files

Weaknesses

  • Only works inside Wherobots Cloud, so your data has to be reachable from its catalogs and storage integrations
  • Public preview: quotas, features and the free-token arrangement may change before general availability
  • Compute is billed for everything it runs, including scheduled runs, and per-chat cost totals are not shown in the chat yet
  • Prompts can be processed in any AWS commercial region under the default model setup
  • No documented API or MCP access to Waldo itself, and it cannot run notebooks or work with local code

Waldo by Wherobots pricing

Professional (trial)

$0

  • 14-day free trial for up to $95 in usage
  • No credit card required
  • Waldo included, with a one-time quota of 55 million AI tokens until a payment method is added

Professional

$95 monthly usage minimum

Billed monthly

  • $95 a month in prepaid usage credits that expire after 90 days
  • Up to 10 users, SSO, MFA and audit logs
  • Waldo AI tokens free during the preview, up to 140 million tokens per week
  • WherobotsDB compute at $1.50 per Spatial Unit in AWS US East and US West

Innovation

From $3,000/month

Billed annual

  • 12-month+ contract with $3,000 a month in usage credits
  • Up to 50 users and premium support
  • Bring Your Own Cloud compute plane in your own VPC

Enterprise

Custom

  • Volume-based pricing and unlimited users
  • 99.5% SLA on Wherobots Cloud compute availability
  • 24x7 enterprise support

Free tier limits: The Professional trial runs for 14 days or up to $95 in usage, whichever comes first; if no payment method is on file when it ends, compute is paused. Trial organizations and organizations without a payment method get a one-time Waldo allowance of 55 million tokens; with a payment method the allowance is 140 million tokens, reset weekly on Mondays at 00:00 UTC. Web search is limited to 2,000 searches per organization per month.

Note: Waldo has no separate price during the public preview: its AI model tokens are included at no additional cost, and you pay only for the WherobotsDB and RasterFlow compute it runs. WherobotsDB is billed per Spatial Unit, roughly the capacity of a 32-vCPU Sedona cluster for one hour ($1.50 in AWS US East and US West, $1.67 in EU West (Ireland), $1.58 in AP South (Mumbai)). RasterFlow tasks are billed per RasterFlow Spatial Unit, calculated from billions of pixels, bands, time periods and a per-task complexity factor ($1.50 for inference, $0.75 for mosaic generation, $0.25 for multiscales and vectorization, minimum 1 SU per task), and Wherobots says RasterFlow pricing may change during or after its preview. Users who found Waldo on Product Hunt can request an extra $250 in credits through a form on the Waldo page.

Technical specs

Modalities

Text, Vector data, Raster imagery, Maps

Available models

OpenAI and Anthropic models served through Amazon Bedrock (specific models not named)

Where Waldo by Wherobots excels

Catastrophe exposure analysis

Waldo can load hurricane track records, build wind footprints, count the buildings and power assets inside them across a whole state and map the result, then schedule it to refresh monthly as versioned tables.

Object detection from aerial imagery

Finding features such as solar panels or rooftops across a city needs imagery inference plus joins to parcel data; Waldo can run RasterFlow inference and spatial SQL in one conversation and map the output for inspection.

Recurring fleet or change-detection pipelines

Analyses that must refresh when new GPS data or satellite imagery arrives can be turned into approved, scheduled pipelines that write results back to an Iceberg lakehouse for other tools to query.

Waldo by Wherobots vs. competitors

Waldo by Wherobots vs. Databricks Genie

Databricks Genie is a governed data agent for business questions on tabular data in Databricks, writing read-only SQL under Unity Catalog permissions. Waldo is aimed at spatial data engineering: it can read Unity Catalog tables through federation, but its strength is spatial SQL, satellite and aerial imagery inference, maps and pipelines that write new data products back to the lakehouse.

Waldo by Wherobots vs. Jedify

Jedify is a business-metrics data agent that monitors KPIs and investigates why they moved, grounded in a company context graph. Waldo does not monitor business metrics; it answers questions about places and builds spatial data products from vector data and imagery.

Waldo by Wherobots vs. Claude Code

Claude Code is a general coding agent in your terminal or editor; connected to the Wherobots MCP server, CLI and agent skills, it can browse Wherobots catalogs, run read-only spatial SQL and submit jobs while also working in your local code and other tools. Waldo is Wherobots' built-in agent: no setup, maps and approvals in the web app, prompts kept in Wherobots' AWS environment, and AI tokens included during the preview instead of paid to a separate model provider.

Frequently asked questions

What is Waldo by Wherobots?

Waldo is a geospatial AI agent for data science and engineering in Wherobots Cloud. You ask a question about places in plain English, and Waldo finds the datasets in your catalogs, writes and runs spatial SQL, Python and RasterFlow imagery tasks on Wherobots, maps the result, and can schedule the work as a data pipeline. It is named after the geographer Waldo Tobler and is unrelated to Glean's Waldo model.

How much does Waldo cost?

During the public preview, Waldo's AI tokens are included at no additional cost, within quotas of 55 million tokens one-time for trial organizations and 140 million tokens per week once a payment method is added. You pay for the WherobotsDB and RasterFlow compute it runs. The Professional plan starts with a 14-day trial for up to $95 in usage with no credit card, then has a $95 monthly usage minimum paid as prepaid credits; WherobotsDB compute costs $1.50 per Spatial Unit in AWS US East and US West.

What data can Waldo work with?

Waldo works with the tables in the Wherobots catalogs your account can query, including Wherobots Open Data such as Overture Maps, your own tables, and catalogs connected through data federation such as AWS Glue, Databricks Unity Catalog and Amazon S3 Tables. It handles vector formats such as GeoParquet, GeoJSON and Shapefile and raster formats such as GeoTIFF, COG and Zarr, with integrations for Iceberg, ESRI and STAC.

Does Waldo ask before it runs jobs or changes data?

Yes. Waldo asks for approval before it starts or stops runtimes, runs jobs, writes or deletes data, creates a Python file or changes an existing file, saves maps, runs billed RasterFlow tasks, or creates or changes schedules. Before a job, it shows the runtime size and hourly cost range. You can approve, always allow, deny or always deny each tool. Read-only actions such as listing tables, reading data and web searches, and by default creating new notebooks or non-Python files, do not need approval.

Which AI models does Waldo use, and is my data used for training?

Wherobots says Waldo uses models from OpenAI and Anthropic, served by Amazon Bedrock in the Wherobots AWS account, and does not name the specific models. Wherobots does not train or fine-tune models on your prompts, chat history, query results or files, and Bedrock does not store or use prompts and responses for training. The default model uses a Bedrock global inference profile, so a prompt can be processed in any AWS commercial region.

Can Waldo turn an analysis into a recurring pipeline?

Yes. Waldo writes a Python script, creates a schedule in a paused state, tests one time window and resumes the schedule after you approve the results. Runs must be at least 15 minutes apart, are billed at normal compute rates, and use your identity and permissions. RasterFlow tasks cannot be scheduled directly, and a schedule is auto-disabled after 5 consecutive failed runs.

Integrations & fit

Wherobots CloudWherobotsDBRasterFlowHavasu CatalogWherobots MapApache IcebergAmazon S3Amazon S3 TablesAWS Glue Data CatalogDatabricks Unity CatalogESRISTACOverture MapsAmazon Bedrock
Good fit forSolo / individual, Startup / small team, Enterprise
Pricing modelPaid· Paid subscription required
See pricing on Waldo by Wherobots →

Alternatives to consider

About Waldo by Wherobots

Waldo is an AI agent for spatial data engineering, data science and analytics from Wherobots, the company founded in 2022 by the original creators of the Apache Sedona spatial engine. It is not related to Glean's Waldo model. It lives inside Wherobots Cloud as an Ask Waldo sidebar on every page and a dedicated chat page. A typical session starts with a prompt that names a place, a measurement and the output you want, such as counting buildings and power assets inside historical hurricane wind footprints in Florida. Waldo asks up to three clarifying questions with suggested defaults, searches the Havasu Catalog and connected catalogs (Databricks Unity Catalog, AWS Glue, Amazon S3 Tables) for the right tables, checks their schemas, then writes and runs SQL or Python on WherobotsDB and, for imagery work, RasterFlow, Wherobots' inference engine for satellite and aerial imagery. Answers come back with a summary, tables, map cards you can open in Wherobots Map, and the methodology and queries it ran. When an analysis should repeat, Waldo writes a Python script, creates a schedule in a paused state, tests one time window and resumes it only when you approve, writing versioned tables back to your Iceberg-compatible lakehouse. What sets it apart from general coding agents is that it works with the catalogs, spatial functions and compute engines directly, under each user's own data permissions, and asks before it starts runtimes, runs jobs, writes data, runs billed RasterFlow tasks or changes schedules. It uses OpenAI and Anthropic models served through Amazon Bedrock in Wherobots' own AWS account rather than the model makers' APIs, and Wherobots says it does not train models on your prompts or data. Waldo launched in public preview on November 18, 2026. During the preview its AI tokens are free within weekly quotas, and you pay only for the WherobotsDB and RasterFlow compute it runs. The tradeoffs: it works only on data reachable from Wherobots Cloud, it is not a desktop GIS replacement, quotas and features may change before general availability, and per-chat compute totals are not shown in the chat yet.

Updates from Waldo by Wherobots

LaunchWaldo launches in public preview

Wherobots released Waldo, its geospatial AI agent for Wherobots Cloud, in public preview on November 18, 2026. It is available in every Wherobots organization edition, including the free Professional Edition trial, with AI tokens included at no extra cost during the preview and compute billed at normal rates. Users who found Waldo on Product Hunt can request an extra $250 in credits through a form on the Waldo page.

Are you the founder? Claim this listing →