Give the question the right amount of work.
Webhound takes two inputs: a research task and an effort budget. The task defines what you need. The budget defines how much searching, reading, and verification Webhound can do before it produces the result.
Choose an output
Start with a Report or Dataset. Choose the shape that matches the answer you need.
Report
A conclusion you will read
Choose a Report for analysis, comparison, explanation, or a recommendation. You receive a cited document plus the sources, claims, and working documents behind it.
Dataset
Rows you will sort or reuse
Choose a Dataset for a market map, company list, product catalog, or lead set. You receive a table with a source trail for each value Webhound finds.
Chains connect Reports and Datasets into a sequence. Ask answers questions from research in your workspace. Use them after you have a first result to build on.
Write the brief
Give Webhound the target, boundaries, deliverable, and proof standard. Name exact locations, dates, counts, or source types when they matter.
- Target: what Webhound must find or decide.
- Boundaries: geography, dates, exclusions, and minimums.
- Deliverable: the structure you need at the end.
- Proof: which sources count and what Webhound must verify.
Report brief
“Compare five US manufacturers that can supply 10,000 custom aluminum bottles per month. Verify minimum order, lead time, certifications, and evidence of capacity. End with a recommendation for a first production run.”
Dataset brief
“Build a dataset of US manufacturers that make custom aluminum bottles. Include location, minimum order, lead time, certifications, capacity evidence, and source URLs. Find at least 25 and mark each value Webhound cannot verify.”
Attach files or past Webhound sessions when they contain useful context. Tell Webhound which facts to use, question, or update.
Set the effort budget
The budget forms part of the task. It gives Webhound room to search, read, compare, verify, and assemble. Use $1 as a 15-minute planning estimate. Webhound starts new Reports and Datasets at $5 unless you change the default.
$2
Map the territory or test a narrow brief.
$5
Run a normal cited investigation.
$10+
Support a decision that needs wider coverage or stronger verification.
Raise the budget for broader coverage, buried details, disputed claims, or a stronger comparison. Source availability and task complexity affect runtime. See Pricing for current minimums and costs.
Start the run
Plan
Confirm the direction first
Plan asks a few questions before a Report starts. For a Dataset, it lets you review the schema and search plan. Leave Plan selected when your brief leaves room for interpretation.
One-Shot
Begin from the brief
One-Shot starts the work when you press send. Choose it when your target, boundaries, deliverable, and proof standard give Webhound enough direction.
A healthy run can keep working after it creates partial documents or rows. Let it finish before you treat the result as complete.
- You can leave the page. The session keeps running in your workspace.
- Answer a checkpoint. Webhound resumes after it receives the missing choice.
- Change the instruction when your goal changes. Send the new scope, constraint, or deliverable in the session chat.
- Add budget when the work uncovers a useful lead. State what the extra effort should investigate.
Use the result
Read the final document or table first. Then inspect the evidence behind the parts that affect your decision.
Inside a Report
- Document holds the final report and working documents.
- Sources lists the pages Webhound used.
- Claims ties factual statements to evidence and confidence.
Inside a Dataset
- Data holds the rows and source trail for each value.
- Sources shows which pages filled the dataset.
- Schema lets you inspect or change the columns.
Open a cited sentence or dataset cell to inspect its evidence. Review weak claims, missing values, and narrow source coverage before you act on the result. Export the final output when you need it in a document, spreadsheet, or another system.
Tips
Useful things to know
Start from a spreadsheet
Choose Dataset, open the + menu, and select Start from spreadsheet. Webhound keeps your rows and columns and can add fields, fill gaps, and verify values against the web.
Name the sources you want checked
Webhound can read public web pages and PDFs. It has dedicated access for sources such as LinkedIn, Reddit, X, YouTube, Google Maps, Amazon, and Crunchbase. A login, paywall, or bot wall can block a page; attach the file or provide API access instead.
Connect services with API keys
Save credentials under Account → Secrets. Webhound can use them during a run to call Notion, Slack, Dropbox, a database, or your own API.
Let Webhound run code
Webhound can write and run Python for calculations, deduplication, parsing, charts, file creation, and API calls. Ask for the result; you do not need to write the code.
Steer the run as it works
Send a new constraint or priority in the session chat. The planner uses it on the next cycle. Add a checkpoint when you want Webhound to pause, show what it found, and wait for your answer.
Reuse work you already have
Attach a past session or folder to use it as starting context for a new run. Use Ask to question a finished result. Start another Report or Dataset when the question requires new evidence.
Use Webhound from your agent
Hand the same task and budget to Codex, Claude, Cursor, or another agent.
Connect Webhound through MCP. Your agent can start a Report or Dataset, wait for completion, and read the final output with its claims, sources, and working documents. Use the API when you need the same workflow inside your product.