Read/write Stardex (recruiting ATS/CRM) — search/get/create/update people (candidates), promote a person to a client contact and change its status or owners, search/get/create/update companies and jobs, list a job's candidates and move a candidate's pipeline stage, add notes to people, jobs and companies, and search/create/update tasks. Use when the user mentions Stardex, pushing candidates into the ATS, or reading jobs, companies, contacts or a pipeline. Keys: BYOK-only — the workspace connects its own Stardex API key.
Stardex is a recruiting CRM, and one of the 56 tools hyreflow connects to Claude, ChatGPT and other AI assistants. Through hyreflow, an AI agent can run 38 actions: 22 that look something up and 16 that create or change records. Stardex runs on your own Stardex account: hyreflow charges no credits for it, and Stardex bills you as before.
You connect hyreflow once over MCP, and Stardex comes with it, next to the other 55 tools. You do not need to run a separate Stardex MCP server.
In Claude, ChatGPT or any other MCP client, add a custom connector named hyreflow with the server URL https://recruit.hyreflow.ai/api/v2/mcp. The MCP setup guide on hyreflow.ai has the exact steps for each app.
A browser window opens. Sign in, choose the workspace the assistant may act on, and approve. The connection is live.
Stardex runs on your own Stardex account. Connect it on the Integrations page in hyreflow, and hyreflow charges no credits for its calls.
Tell Claude or ChatGPT what you want done with Stardex in plain language. The assistant picks the Stardex action that fits and runs it through hyreflow.
Using Cursor, Claude Code or another client? Every one is covered in the MCP setup guide.
Through hyreflow, an AI agent can run 38 Stardex actions: 22 that look something up and 16 that create or change records. Stardex runs on your own Stardex account: hyreflow charges no credits for it, and Stardex bills you as before.
| Stardex action | What it does | Cost through hyreflow |
|---|---|---|
| Search candidates | Looks something up | no hyreflow credits |
| Search companies | Looks something up | no hyreflow credits |
| Search jobs | Looks something up | no hyreflow credits |
| Search persons | Looks something up | no hyreflow credits |
| Search tasks | Looks something up | no hyreflow credits |
| Create client contact | Creates or changes a record | no hyreflow credits |
| Create company | Creates or changes a record | no hyreflow credits |
| Create company activity | Creates or changes a record | no hyreflow credits |
| Create job | Creates or changes a record | no hyreflow credits |
| Create job activity | Creates or changes a record | no hyreflow credits |
| Stardex action | What it does | Cost through hyreflow |
|---|---|---|
| Create people batch | Creates or changes a record | no hyreflow credits |
| Create person | Creates or changes a record | no hyreflow credits |
| Create person activity | Creates or changes a record | no hyreflow credits |
| Create task | Creates or changes a record | no hyreflow credits |
| Update candidate stage | Creates or changes a record | no hyreflow credits |
| Update candidate stage batch | Creates or changes a record | no hyreflow credits |
| Update client contact | Creates or changes a record | no hyreflow credits |
| Update company | Creates or changes a record | no hyreflow credits |
| Update job | Creates or changes a record | no hyreflow credits |
| Update person | Creates or changes a record | no hyreflow credits |
| Update task | Creates or changes a record | no hyreflow credits |
| List client contact statuses | Looks something up | no hyreflow credits |
| List client contacts | Looks something up | no hyreflow credits |
| List company activity types | Looks something up | no hyreflow credits |
| List company statuses | Looks something up | no hyreflow credits |
| List custom fields | Looks something up | no hyreflow credits |
| List job candidates | Looks something up | no hyreflow credits |
| List job statuses | Looks something up | no hyreflow credits |
| List person activities | Looks something up | no hyreflow credits |
| List person activity types | Looks something up | no hyreflow credits |
| List pipeline stages | Looks something up | no hyreflow credits |
| List rejection reasons | Looks something up | no hyreflow credits |
| List team members | Looks something up | no hyreflow credits |
| Get candidate | Looks something up | no hyreflow credits |
| Get client contact | Looks something up | no hyreflow credits |
| Get company | Looks something up | no hyreflow credits |
| Get job | Looks something up | no hyreflow credits |
| Get person | Looks something up | no hyreflow credits |
One credit is priced on the pricing page. The full price list for every tool is in the pricing tables.
Ask for the outcome and your agent composes the run itself. Or start from a skill, a whole pipeline it already knows end to end.
browse all skills →hyreflow ships a written playbook for Stardex, and the agent loads it before the first call. Auth, rate limits, what each call costs, which actions need your approval, and the mistakes worth avoiding. It is the difference between an agent that knows the tool and one that guesses at it.
Read the Stardex docs →HOW in lib/stardex.py. Docs: docs.stardex.com/api-reference/introduction-v1 (published OpenAPI spec).
hyreflow byok set stardex --api-key <key>.403 INSUFFICIENT_SCOPE. Hyreflow needs no Delete and no Enroll & send.team_member_id (person notes only; company and job notes always have none) or owner_ids (tasks): find ids with list_team_members.RATE_LIMIT_EXCEEDED 429. List and search calls return one page: offset from 0, limit at most 100 (200 on search_jobs); a larger value is rejected, not trimmed. Keep raising offset until it reaches meta.total.Person = any human record, and the candidate record itself. Client contact = a person promoted onto a client company (it keeps its person id). Company = an account; a client once it has a client contact or is_client. Job = a role with its own pipeline of stages. Candidate = one person on one job's pipeline, with its own candidate_id. Activity = a note or other logged event (secondary_note is the general note).
People: read search_persons / get_person / list_person_activities / list_person_activity_types; write create_person, update_person, create_person_activity (a note).
Client contacts: read list_client_contacts / get_client_contact / list_client_contact_statuses; write create_client_contact, update_client_contact.
Companies: read search_companies / get_company / list_company_statuses / list_company_activity_types; write create_company, update_company, create_company_activity.
Jobs and pipeline: read search_jobs / get_job / list_job_candidates / list_job_statuses / list_pipeline_stages, search_candidates / get_candidate / list_rejection_reasons; write create_job, update_job, create_job_activity, update_candidate_stage.
Tasks and workspace: search_tasks, create_task, update_task, list_team_members, list_custom_fields (persons, companies or jobs).
Bulk: create_people_batch, update_candidate_stage_batch, plus the generic batch for get/search/update/create calls (see crm-bulk-pushes.md).
create_person may not create a new record. With linkedin_url, Stardex matches the existing person with that profile and updates them; without it, a person with the same name who also shares an email or phone can be merged into the existing record. Either way you get the existing person's id back. Always send linkedin_url when you have it, and use update_person once you hold an id. Only name is required: {"name": "Jane Doe", "linkedin_url": "https://www.linkedin.com/in/jane-doe", "work_email": "[email protected]", "current_job_title": "Staff Engineer", "company_name": "Acme"}.job_id on create_person or update_person; that creates the candidacy. Then get_job lists the job's pipeline_stages, and update_candidate_stage moves them: {"candidate_id": "<id>", "stage_id": "<stage id>"}, or job_id plus candidate_email when you have no candidate id (the job is checked first, since Stardex would otherwise move the person on another job). Prefer ids: Stardex matches stage_name and candidate_name loosely. rejection_reason_id and reason_text are only allowed on a rejection stage. A stage move starts the workspace's Stardex automations that are set to run on API changes, so check with the user before a bulk move.update_person: only the fields you send change. work_email, personal_email, mobile_phone and work_phone are added to the existing list; emails / phone_numbers replace the whole list and can't be sent with the single fields. A linkedin_url another person already has answers 409 LINKEDIN_URL_IN_USE.create_person first, then create_client_contact(person_id, client_id). A person can be the client contact of one company only: if they already are, you get the existing record back with is_new: false and its real client_id. update_client_contact changes only the status and owners (owner_ids replaces all of them); change names and emails with update_person.create_company returns a match when there is one (by LinkedIn URL or domain, or exact name for a company with neither) and then ignores most of the fields you sent. Read the returned company and use update_company to change it. It needs at least one of name, domain, linkedin_url.create_job needs title and one of company_name, company_domain, company_linkedin_url. On update_job, deal_ids and client_contact_ids replace every existing link.create_person_activity finds the person by person_id (best), email, linkedin_url or full_name; create_job_activity and create_company_activity take the record id. activity_type defaults to secondary_note. Notes in HTML can @mention a teammate, which emails them.task_type is one of text, linkedin_message, linkedin_connection, call, other.ATS write target: enriched candidates → create_people_batch (with linkedin_url and job_id on each row) → update_candidate_stage as they progress; reply or call notes → create_person_activity.
Company-first BD: search_companies → each company's website → hyreflow_native career-page scraping for their open roles → contacts via the people_search waterfall → create_person + create_client_contact.
For large pushes use create_people_batch. It paces calls to the 50 req/min default and returns
{created, failed, total, elapsed_ms}. See crm-bulk-pushes.md for the cross-CRM guide.
Call via the CLI: hyreflow tools execute stardex <method> --payload '{...}' (preview with --dry-run; hyreflow tools get stardex <method> returns the live contract + cost). Base: https://api.stardex.ai/v1. Only the methods listed below are callable.
batch(method: str, calls: list, *, stop_on_error: bool = False) -> dict — Batched get/search/update/create: run one method over many inputs, rate-limited. Each call is a value -> method(value), or {"args": [...], "kwargs": {...}} for multi-arg methods like update_person(id, payload). Destructive methods (delete) are refused. See lib/_base.run_ops.create_client_contact(person_id: str, client_id: str, *, client_contact_status_id: str | None = None, owner_ids: list[str] | None = None) -> Any — POST /client-contacts — promote an existing person (create_person first) to a client contact of client_id, which becomes a client. Idempotent: a person already a client contact comes back with is_new false and their actual client_id. owner_ids [] creates it with no owners.create_company(payload: dict) -> Any — POST /companies — create a company, or return the existing one it matches (by LinkedIn URL or domain; by exact name only for companies with neither). At least one of name, domain, linkedin_url; also description, location, phone, is_client, custom_fields[]. On a match most fields are NOT applied: use update_company on the returned id to change it.{"name":"Acme","domain":"acme.com","linkedin_url":"https://www.linkedin.com/company/acme","is_client":true}create_company_activity(company_id: str, activity_content: str, *, activity_type: str = 'secondary_note') -> Any — POST /company-activities — log a note on a company. activity_type must be one of the team's company activity types (list_company_activity_types); secondary_note is the general note.create_job(payload: dict) -> Any — POST /jobs — create a job. title and one of company_name, company_domain, company_linkedin_url (the client, matched or created) are required; also job_description, location, primary_url, currency_code, job_function, open_date / end_date (YYYY-MM-DD), salary_min / salary_max or salary_range, salary_details, fee, firm_commission_percent, pipeline_template_name, team_members[], job_status_id, deal_ids[] and client_contact_ids[] (max 50 each), billing_contact_name / billing_contact_email.{"title":"Staff Engineer","company_domain":"acme.com","location":"London, UK","job_description":"Lead the platform team.","salary_range":"120000-150000","currency_code":"GBP"}create_job_activity(job_id: str, activity_content: str, *, activity_type: str = 'secondary_note') -> Any — POST /job-activities — log a note (secondary_note, call_note, meeting_note, email, ...) on a job.create_people_batch(payloads: list[dict], *, stop_on_error: bool = False) -> dict — POST /persons for many records, one at a time, respecting the Stardex rate limit. Same match/merge rules as create_person: send linkedin_url on every row you have it for.create_person(payload: dict) -> Any — POST /persons — create a person. name is required; also first_name, last_name, linkedin_url, current_job_title, location, company_name, work_email, personal_email, emails[], work_phone, mobile_phone, phone_numbers[], notes, job_id (adds them to that job's pipeline as a candidate), client_id, external_source_id, enrich_record (Stardex defaults it to true; this adapter sends false unless you set it, so a push does not start Stardex's own enrichment), person_owner, sourced_by (needs job_id), custom_fields[].{"name":"Ada Lovelace","linkedin_url":"https://www.linkedin.com/in/ada-lovelace","current_job_title":"Staff Engineer","company_name":"Acme","work_email":"[email protected]","location":"London, UK","job_id":"9b2f4c1e-3d5a-4e6b-8c7d-1a2b3c4d5e6f"}create_person_activity(activity_content: str, *, person_id: str | None = None, email: str | None = None, linkedin_url: str | None = None, full_name: str | None = None, activity_type: str = 'secondary_note', team_member_id: str | None = None, job_id: str | None = None, occurred_at: str | None = None) -> Any — POST /person-activities — log a note (default type secondary_note, the general note) or another activity (meeting_note, call_note, email, interview_note, ... see list_person_activity_types) on a person, found by one of person_id (most precise), email, linkedin_url or full_name. job_id ties it to a job the person is already a candidate on; team_member_id records the author (none otherwise); occurred_at backdates it (ISO 8601).create_task(payload: dict) -> Any — POST /tasks — create a task. content_text and task_type (text, linkedin_message, linkedin_connection, call, other) are required; also due_date (ISO 8601), is_completed, person_id, client_id, job_id, deal_id, owner_ids[] (ownerless when omitted on an API key).{"content_text":"Call Ada about the Staff Engineer role","task_type":"call","due_date":"2026-11-20T10:00:00Z","person_id":"2f6b9d1e-4a8c-4e3f-b7d2-1c5a9e3b7f40"}get_candidate(candidate_id: str) -> Any — GET /candidates/{id} — one candidate with their full stage history.get_client_contact(client_contact_id: str) -> Any — GET /client-contacts/{id} — one client contact; a person id is accepted too.get_company(company_id: str, *, include: list[str] | str | None = None) -> Any — GET /companies/{id} — one company with client status, owners and custom fields.get_job(job_id: str, *, include: list[str] | str | None = None) -> Any — GET /jobs/{id} — one job with its pipeline_stages (stage ids for update_candidate_stage), team members, candidate count and custom fields.get_person(person_id: str, *, include: list[str] | str | None = None) -> Any — GET /persons/{id} — one person. include picks sections (work_experience, education_experience, certifications, candidates, custom_fields, activities, compensation, references, offers, team_connections); default work_experience, education_experience, certifications, custom_fields.list_client_contact_statuses() -> Any — GET /client-contacts/statuses — the client contact statuses (ids for client_contact_status_id).list_client_contacts(*, limit: int = 100, offset: int = 0, **filters) -> Any — GET /client-contacts — one page of client contacts with their person, company and status. Filters: person_id, client_id, client_contact_status_id, keywords (name contains), is_referral_partner, sort_by (created_at, updated_at), sort_order (asc, desc).list_company_activity_types() -> Any — GET /company-activities/types — the activity type keys create_company_activity accepts.list_company_statuses() -> Any — GET /companies/statuses — the company client statuses (ids for client_status_id).list_custom_fields(entity: str, **filters) -> Any — GET /custom-fields/{persons|companies|jobs} — custom field definitions (attribute ids and tag options for custom_fields[] and the *_attribute_filters). persons/companies take job_id and list_id to add job- or list-specific fields; companies also takes is_client.list_job_candidates(job_id: str, *, limit: int = 100, offset: int = 0) -> Any — GET /jobs/{id}/candidates — every candidate on one job, unfiltered (search_candidates filters).list_job_statuses() -> Any — GET /jobs/statuses — the job statuses (ids for job_status_id) with their lifecycle category.list_person_activities(person_id: str, *, limit: int = 100, offset: int = 0, **filters) -> Any — GET /persons/{id}/activities — the person's timeline (notes, emails, tasks, meetings, stage changes), newest first. Filters: search, team_member_ids[], job_ids[], activity_types[], excluded_activity_types[], pipeline_stage_ids[], created_at_gte, created_at_lte.list_person_activity_types(*, include_archived: bool | None = None) -> Any — GET /person-activities/types — every activity type key for the team, built-in and custom.list_pipeline_stages(*, include_deleted: bool | None = None) -> Any — GET /jobs/pipeline-stages — the team-wide stage catalog. A job's own stages and order: get_job.list_rejection_reasons() -> Any — GET /candidates/rejection-reasons — the active rejection reasons (ids for rejection_reason_id).list_team_members(*, limit: int = 100, offset: int = 0, **filters) -> Any — GET /team-members — the recruiters (ids for owners, sourced_by, team_member_id). Filters: is_active, first_name, last_name, email_address.search_candidates(*, limit: int = 100, offset: int = 0, **filters) -> Any — POST /candidates/search — candidates across jobs, one row per person and job (a read; limit <= 100).search_companies(*, limit: int = 100, offset: int = 0, **filters) -> Any — POST /companies/search — find companies (a read; limit <= 100). Filters: keywords[] (name, domain, LinkedIn), domain (exact), linkedin_url (exact), is_client, company_ids[], company_status_ids_included, owner_ids_included, job_id, job_association (has_jobs, no_jobs); include ["custom_fields"].search_jobs(*, limit: int = 100, offset: int = 0, **filters) -> Any — POST /jobs/search — find jobs (a read; limit <= 200). Prefer vector_search (5+ chars) for a role by title. Filters: keywords[], location_keywords[], is_archived, job_status_ids_included, company_ids_included, owner_ids, team_member_ids, created_at_gte/lte, updated_at_gte/lte; include[] (custom_fields, pipeline_stages, client_contacts, ...).search_persons(*, limit: int = 100, offset: int = 0, **filters) -> Any — POST /persons/search — find and filter persons (a read; limit <= 100). Direct lookups: email (exact), name (ranked, top 50), phone, linkedin_url (exact), vector_search (semantic, 5+ chars), keywords[].search_tasks(*, limit: int = 100, offset: int = 0, **filters) -> Any — POST /tasks/search — tasks (a read; limit <= 100). Filters: keywords[], task_types[], related_to + related_record_type (person, deal, client, job; always together), is_completed, owner_ids[], due_date_start, due_date_end; sort_by (created_at, updated_at, due_date).update_candidate_stage(payload: dict) -> Any — POST /candidates/update-stage — move a candidate to another stage of their job. Name the candidate by candidate_id, or by job_id plus candidate_email / candidate_name (must match exactly one person); name the stage by stage_id (from get_job's pipeline_stages) or stage_name. Prefer ids: Stardex resolves names loosely. rejection_reason_id (list_rejection_reasons) and reason_text are only allowed on a rejection stage. Without candidate_id the job_id is checked first (get_job): Stardex treats an unknown job_id as absent and would move the person's most recent candidacy on another job instead.{"candidate_id":"789e0123-e45f-67a8-b901-234567890123","stage_id":"4c2a8e1f-6b3d-4f9a-a7c5-2e8d1b6f3a90"}update_candidate_stage_batch(payloads: list[dict], *, stop_on_error: bool = False) -> dict — POST /candidates/update-stage for many candidates, rate-limited. Each payload is what update_candidate_stage takes.update_client_contact(client_contact_id: str, payload: dict) -> Any — PATCH /client-contacts/{id} — change client_contact_status_id (null clears) and/or owner_ids (replaces every owner; [] removes all). A person id is accepted too. Other person fields go through update_person.update_company(company_id: str, payload: dict) -> Any — PATCH /companies/{id} — partial update: name, domain, linkedin_url, description, location, phone, is_client, client_status_id (null clears), owner_ids (replaces all), do_not_contact, do_not_contact_reason, do_not_contact_expiration_date, custom_fields[].update_job(job_id: str, payload: dict) -> Any — PATCH /jobs/{id} — partial update: title, job_description, location, job_status_id, salary fields, fee, dates, is_archived, custom_fields[]. deal_ids / client_contact_ids REPLACE every existing link.update_person(person_id: str, payload: dict) -> Any — PATCH /persons/{id} — partial update; omitted fields are kept. Fields: name, first_name, last_name, linkedin_url (409 LINKEDIN_URL_IN_USE if another person has it), current_job_title, company_name, location, linkedin_headline, address, work_email / personal_email / mobile_phone / work_phone (added to the existing contacts), emails[] / phone_numbers[] (replace the whole list; not combinable with the single fields), job_id (adds them to that job's pipeline), sourced_by, person_owner (null clears), do_not_contact, do_not_contact_reason, do_not_contact_expiration_date, custom_fields[], enrich_record.update_task(task_id: str, payload: dict) -> Any — PATCH /tasks/{id} — change content_text, task_type, due_date (null clears) or is_completed.Yes. hyreflow is a connector for Claude, ChatGPT and other AI assistants. Once hyreflow is connected, you ask for the outcome in plain language and the assistant calls Stardex for you. hyreflow exposes 38 Stardex actions this way.
Yes. Stardex runs on your own Stardex account. You connect it to hyreflow once, hyreflow charges no credits for the calls, and Stardex bills you as before.
Stardex runs on your own Stardex account: hyreflow charges no credits for it, and Stardex bills you as before.
Add the hyreflow MCP server to your assistant and sign in. hyreflow calls the Stardex API directly and exposes its actions as tools the assistant can run, so there is no separate Stardex server to install.
Connect hyreflow to Claude or ChatGPT and ask for the outcome. It reads the playbook above before it makes a single call.