Glasser
Why we built Glasser
Give your agent access to paid data with one key, without a separate subscription for each provider.
You ask your agent to research a list of companies. You want their websites, where they operate, and enough evidence to decide which ones deserve a closer look. Halfway through, you need data from a paid provider. Now you have to create an account, choose a plan, and configure another key before the agent can continue.
A second task may need a different provider. You repeat the setup, even if you only need a handful of records. We built Glasser to reduce that work: one key for paid data, with no monthly commitment.
Get the data you need for this task
An agent can help you write a research plan and compare the records you give it. To check a company detail or retrieve search results, it needs access to data beyond the conversation. You still have to arrange that access.
With Glasser, you can use the endpoints in Data Sources through one key. We manage the provider credentials. Your agent can find an endpoint for the task and use it without asking you to register with that provider.
You might need company data for a shortlist this week and web search for a different project next week. You can use either through the same setup, and pay according to the pricing rule of each endpoint you run.
Search, inspect, then run
Glasser gives your agent a way to discover endpoints while it works. It searches the data sources, inspects a candidate endpoint, and runs it with the required inputs. You can use the CLI, MCP, or API to reach the same endpoints.
Inspection matters. Two endpoints that mention companies may answer different questions: one finds matches, while another retrieves details for a company you already know. Your agent can check the inputs and pricing before choosing which one to use.
For a company shortlist, start with a clear brief: the region, the type of business, and the fields you need. Have the agent retrieve a small set of candidates and check the response before requesting more. If a field is missing, keep that gap visible in the shortlist.
Check the price before you run it
A small research project should be something you can try without choosing a monthly data plan. Glasser uses a prepaid balance. Each endpoint has a published pricing rule, including its billable unit and maximum charge.
Some endpoints charge per call; others charge per result. Your agent can inspect that distinction before it runs an endpoint, then check the actual charge afterward. You can judge a workflow using the data you received and what you paid for it.
Keep the records behind the answer
When your agent hands you a shortlist, you should be able to check why a company made the cut. Glasser returns the provider's response, with sensitive fields redacted where needed. Your agent uses those records to build its answer.
Each run also has a Run URL where workspace members can review the retrieved data. Ask your agent to include it with the output so you can return to the records when you need to check a detail.
Access to paid data does not make every record complete or every conclusion correct. You and your agent still need to check dates, missing fields, and whether the evidence supports the answer. We want to make that review possible alongside the work.
Bring one question you want answered
Start with a task you can review in one sitting. Ask for a shortlist of companies in a market you know, or research a topic where you can check the sources. Tell your agent what a useful answer must contain and ask it to inspect the relevant endpoints before running them.
Open Data Sources to see what is available, then follow the getting-started guide to connect your agent. Try a small request, read the records, and decide whether they help you take the next step.