Namsor
Analytics & DataClassify personal names with Namsor to infer likely gender, country of origin, country of residence, and diaspora. Every result carries a confidence score, so agents can route uncertain cases to a person.
What This Integration Enables
Namsor infers more from a name than anything else in FlowRunner's catalog, and that is precisely why this page spends its first paragraphs on limits rather than capabilities. The classifiers cover likely gender, country of origin, country of residence, diaspora or ethnicity within a country, and a US Census race and ethnicity taxonomy. Every one of them returns a machine-learning inference with a calibrated probability. None of them returns a fact. A diaspora classification is a statement about how a name distributes across populations, not a person's declared or verified ethnicity, and Namsor's own documentation says so.
The uses that hold up are aggregate and administrative. In aggregate: measuring the composition of an applicant pool, a customer base or a citation dataset at the group level, which is the standard method behind published bias and diversity research. Administratively: Parse Name and Parse Names Batch split unstructured full names into first and last components across formats like "John Smith", "SMITH, John" and "SMITH; John", returning a nameParserType structure code and a confidence, with country context materially improving accuracy on conventions like Spanish two-surname names. Get Name Type separates a personal name from a brand name, a pseudonym or a place name, which is the cheapest way to clean a lead list before anything else touches it. Country of residence and origin support routing a record to the right locale.
The use that does not hold up, and that FlowRunner will not help you build, is attaching an inferred gender, ethnicity, national origin or race to a named individual and letting it influence a decision about that person. Hiring, lending, pricing, insurance, eligibility: an inference from a name has no place in any of them, and in many jurisdictions using it that way is unlawful. The rule the flows on this page follow is simple and worth stating once: inference is for addressing people correctly and for understanding populations, not for deciding about individuals. Namsor also exposes Get API Key Info, which reports the learnable, vetted and anonymized flags governing whether submitted names are retained for machine learning. Reading that before you submit personal data is not optional diligence, it is the first step of the flow.
Without FlowRunner
With FlowRunner
Use Case Scenarios
Cleaning a signup list before anything else runs
An imported lead list contains people, company names, and a few rows where somebody typed a city into the name field. The agent runs Get Name Types Batch across the list, which classifies each entry as anthroponym, brand-name, pseudonym or toponym at 1 credit per name, and splits the list accordingly. The brand rows are routed to company enrichment, the place rows are flagged as bad data, and only the person rows continue. This step costs almost nothing and prevents the expensive classifiers from being pointed at things that are not people, which is both a spend problem and an accuracy problem.
Normalising name structure across an international import
A partner feed arrives with a single name column in inconsistent order and casing. The agent runs Parse Names Batch with the country code already on each record, which is what makes Spanish two-surname names and inverted "SMITH, John" formats resolve correctly. The result carries the detected structure code and a confidence per row. High-confidence parses are written back to Attio as first and last name. Low-confidence parses are left as the original single field rather than being split wrongly, because a badly split name is worse than an unsplit one for every downstream template.
Aggregate composition analysis with the grain fixed up front
A research team measures the composition of a historical citation dataset. The agent parses and type-filters the names, then runs Get Names Origin Batch over the cleaned set and aggregates to region-level counts inside the same flow, before anything is stored. What persists is a distribution with its confidence spread and the share of names that could not be classified. The per-name classifications are not written to any record, not joined back to individual authors, and not retained. The output is a chart about a population, and the flow is built so that it cannot accidentally become a label on a person.
Human-in-Loop Highlight
The gate goes in front of the classifier, not after it, and it exists because the sensitive step here is submission itself. Once a hundred real names have been sent to a diaspora or race classifier, the inference exists, the credits are spent at 20 per name, and if the key's learnable flag is on, the names may be retained for training. None of that can be walked back by discarding the output. So the agent does the cheap, reversible preparation on its own, running Parse Names Batch and Get Name Types Batch, then calls Get API Key Info and Get API Usage and stops. It posts to the analysis owner: "Cohort cleaned to 1,840 person records from 2,100 rows. Key flags: learnable true, anonymized false. Get Names Diaspora Batch over this cohort costs 36,800 credits against a remaining balance of 50,000. Output is configured as country-level aggregate counts with no per-record write. Confirm the retention setting and approve, or change the key first?" Nothing is submitted until a person answers. Note what the agent surfaced without being asked: the retention flag that most integrations would never look at. That is what the digital andon cord is for, stopping the line at the point where continuing becomes irreversible.
Agent Capabilities
17 actionsGender
2- Genderize Name Infers the most likely gender associated with a personal name from either a full name or a first name with an optional last name, with a country of residence improving accuracy on locally ambiguous names. Returns likelyGender with a calibrated probability and a signed genderScale from -1 to +1. The result describes how the name is typically used in the given context and is not a verified fact about any individual. Costs 1 credit per name.
- Genderize Names Batch Infers likely gender for up to 100 personal names in one call at 1 credit per name. Each entry carries a full name or separate first and last names plus an optional ISO 3166-1 alpha-2 country code, and all entries in one call must use the same style. Results come back in the personalNames array with the id you supplied, so they join back to your records. Each result is a calibrated statistical inference, not a verified fact.
Origin
2- Get Name Origin Infers the most likely country of cultural origin of a personal name with a region, sub-region, the top ten candidate countries and a calibrated probability. Costs 10 credits per name. This classifier assumes the name is written as it would be in its country of origin, so for residents of melting-pot countries such as the US, CA, AU or NZ use Get Name Diaspora instead. The output describes a name distribution, not an individual's nationality or birthplace.
- Get Names Origin Batch Infers likely country of cultural origin for up to 100 names in one call at 10 credits per name, with all entries using the same name style. This endpoint takes no country context and assumes names as written in their country of origin. Returns the top ten candidates, region information and calibrated probabilities as probabilistic inferences.
Country of Residence
2- Get Name Country Of Residence Infers the most likely current country of residence associated with a personal name, returning the best and second-best country, the top ten candidates, region information and calibrated probabilities. Costs 10 credits per name. Unlike Get Name Origin, it assumes names as written either in the country of residence or the country of origin. The result is an inference over name distributions and does not establish where any specific person lives.
- Get Names Country Of Residence Batch Infers likely country of residence for up to 100 names in one call at 10 credits per name, with all entries using the same name style. Used to regionalise a contact list or route records to the right locale. Every result carries a calibrated probability and is not a verified address or nationality.
Diaspora
2- Get Name Diaspora Infers the most likely ethnicity or diaspora group of a personal name within a given country of residence, returning the best and second-best ethnicity, the top ten candidates and calibrated probabilities. Costs 20 credits per name. This is the classifier intended for residents of melting-pot countries such as US, CA, AU and NZ. Results are probabilistic inferences with a stated confidence and must never be treated as a person's declared or verified ethnicity, or used to make a decision about an individual.
- Get Names Diaspora Batch Infers likely ethnicity or diaspora group for up to 100 names in one call at 20 credits per name. Every entry must carry a countryIso2 country of residence and use the same name style as the rest of the call. The lifted flag on each result indicates whether the ethnicity was reinforced by a known fact rather than by machine learning alone. All results are probabilistic inferences intended for aggregate analysis, never verified attributes of an individual.
US Race and Ethnicity
2- Get Name US Race Ethnicity Classifies the name of a US resident against the US Census race and ethnicity taxonomy, returning the best and second-best class, the ranked list and calibrated probabilities. Costs 10 credits per name. The default taxonomy has four classes and a six-class option adds AI_AN and PI. Supplying a US ZIP code alongside first and last names sharpens the inference. The output is intended for aggregate analysis such as bias auditing and is not a person's self-identified race or ethnicity.
- Get Names US Race Ethnicity Batch Classifies up to 100 names of US residents against the US Census taxonomy in one call at 10 credits per name, with an optional countryIso2 and, for first and last name entries, a US zipCode for geographic context. All entries in one call must use the same name style. Results are calibrated probabilistic inferences intended for aggregate diversity or bias measurement rather than as facts about individuals.
Name Type
2- Get Name Type Determines the most likely type of a proper noun, distinguishing a personal name from a brand name, a pseudonym or a place name. Costs 1 credit per name. Returns commonType, a second-best commonTypeAlt and a confidence score, which makes it the cheap first pass for filtering non-person records out of a lead or signup list. A country code sharpens the inference for locally specific brands and places.
- Get Name Types Batch Determines the type of up to 100 proper nouns in one call at 1 credit per name, classifying each as anthroponym, brand-name, pseudonym or toponym, with an optional country code per entry. Results come back in the properNouns array with the id you supplied. Used to clean a large signup or CRM list before any person-level classifier runs.
Name Parsing
2- Parse Name Splits an unstructured full name into a first name and a last name, handling formats such as "John Smith", "SMITH, John" and "SMITH; John". Costs 1 credit per name. Returns the split components plus nameParserType, a structure code such as FN1LN1 or LN1FN2, an alternative parse and a confidence score. A country code materially improves accuracy where naming conventions differ, for example Spanish two-surname names.
- Parse Names Batch Splits up to 100 unstructured full names in one call at 1 credit per name, each entry with an optional ISO 3166-1 alpha-2 country code. Results come back in the personalNames array with the id you supplied, the detected structure code and a confidence score. Used to normalise an imported contact list before any further processing.
Account
3- Get API Usage Returns current credit consumption for the connected key: the active plan and quota, usage so far in the billing period, the soft and hard limits, and any overage quantity and amount. Admin routes are free. Used to check spend before a large batch, since origin, country and US race calls cost 10 credits per name and diaspora costs 20.
- Get API Key Info Returns the configuration and privacy settings attached to the connected key: the associated user id, whether the key is disabled, and the learnable, vetted and anonymized flags that control whether submitted names are retained for machine learning. Admin routes are free. Run this before processing personal data, and to check the explainable flag that governs the optional AI explanation output.
- Get API Status Returns the running software version and the operational status of every Namsor classifier, including whether each is serving, still learning, shutting down, or finished calibrating its probabilities. When probabilityCalibrated is false for a classifier, its results report a probability of -1 instead of a calibrated confidence, so this is the check to run when confidence values look wrong.
Frequently Asked Questions
What can FlowRunner do with Namsor?
FlowRunner agents can run Genderize Name, Genderize Names Batch, and Get Name Origin in Namsor, plus 14 more actions.
Does connecting Namsor to FlowRunner require OAuth?
No. Namsor connects to FlowRunner with an API key, no OAuth flow required.
Can Namsor trigger a FlowRunner workflow automatically?
Namsor doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger.
Start building with Namsor
$100 in credits. No card required. Connect in minutes.