CAPABILITY MAP / SEPTEMBER 9, 2026
Our mapping of the Edges capabilities supplied for review: what ProfileGlide implements, what you can compose, and what still needs work. It is not a claim of full catalog or field parity.
Implemented means the named behavior exists in our code. Eight data actions have bounded live-cloud validation; this does not mean the public service is deployed.
Build with API means your application supplies orchestration. Needs implementation is a development path, not an available endpoint. Needs account access requires a suitable customer identity. Not established means the claim lacks evidence or a guaranteed source.
linkedin-extract-people returns identity, headline, summary, experience, education, skills, and counts when supplied.
Boundary / next step: 20 credits per returned profile. No promise of 38 populated fields, every historical role, or private contact data.
salesnavigator-search-person matches name + current company using Sales Navigator search only.
Boundary / next step: 1 credit per returned match. No profile visit; no URL-only lookup or hidden full-profile fallback.
linkedin-extract-company returns company name, website/domain, industry, specialties, locations, size range, and source employee count.
Boundary / next step: 1 credit per company. The source’s employee_count is not independently verified exact payroll headcount.
Search rows expose job_title, company_name, tenure_start. Full profiles expose experiences[].title, company_name, and date.
Boundary / next step: These are different action schemas. Full-profile output does not expose a top-level job_title. Date completeness depends on the source.
Available experience, education, skills, counts, and connection degree are projected.
Boundary / next step: has_more_experiences and has_more_skills can signal truncation. Connection degree is relative to the selected source account. We cannot promise every role or every section.
Search outputs retain source-provided Sales Navigator profile/company references.
Boundary / next step: Full-profile and company reads do not guarantee Sales Navigator IDs. Missing IDs stay missing; a separate resolver would need validation.
Native Unipile company detail is a candidate for richer funding/insights fields after checking Sales Navigator paths.
Boundary / next step: These fields are not in our current company projection. Validate payload coverage, dates, amounts, and units before adding them.
Similar-company discovery can use firmographic Sales Navigator company search. Exact affiliate relations need source evidence.
Boundary / next step: Similarity is not ownership. Exact affiliated-page/recommendation panels may require a validated Unipile raw read after search/native routes are exhausted.
New executions request source data; completed idempotent replays return the original run output.
Boundary / next step: We cannot guarantee that the source itself has no cache, all fields are visible, or every dataset competitor is stale.
salesnavigator-search-people accepts a keyword string and returns up to 100 records from one source page.
Boundary / next step: No natural-language filter compiler. Geography, seniority, title, and headcount are not public request parameters yet.
salesnavigator-search-companies, linkedin-search-companies, and salesnavigator-search-company-employees provide firmographic discovery and company employee lead extraction.
Boundary / next step: Single page results up to 100 records. Multi-page pagination engine is roadmap work.
Unipile documents Sales Navigator list/search references scoped to the connected identity.
Boundary / next step: A customer’s private lists cannot be read through a shared managed account. Requires a tenant-owned identity connection and lifecycle.
Evaluate documented event-related search paths, then native event reads, then fixed raw adapters for remaining fields.
Boundary / next step: Sales Navigator keyword hits do not prove event attendance. Event discovery and a complete attendee roster need separate validation and account visibility.
linkedin-extract-school-alumnis and linkedin-extract-group-members extract search-visible matching people; linkedin-search-schools and linkedin-search-groups resolve entity parameters.
Boundary / next step: Private group roster join timestamps and private member lists require customer identity.
salesnavigator-search-company-metrics returns exact total_results from the search index without charging for full candidate rows.
Boundary / next step: Full department distribution breakdown across all departments is unbundled; single metric query returns total search pool size.
Evaluate Sales Navigator hiring/account search for company signals, then native Unipile job search/detail for actual vacancies.
Boundary / next step: A company hiring flag is not a job record. Rich fields and exact job-detail parity need a tested adapter; raw is last fallback.
Sales Navigator has identity-relative viewer/follower filters for specific cases.
Boundary / next step: They concern the connected identity or its company, not arbitrary profiles/pages. Full viewer/roster access may need page-admin rights; anonymous viewers cannot be unmasked.
linkedin-search-content and linkedin-extract-people-post-activity return text, dates, IDs, and available engagement counts.
Boundary / next step: Profile posts need linkedin_profile_identifier from enrichment. No full activity timeline or company-post action in the public catalog.
Pass a returned social_id to linkedin-extract-post-commenters or linkedin-extract-post-likers.
Boundary / next step: 1 credit per returned comment or reaction, not per page. These are comment/reaction records with available actor context, not fully enriched Profile[] or unique people.
Unipile connected-account operations provide a path to evaluate identity-owned write actions.
Boundary / next step: Managed scraping approvals do not authorize customer outreach. Needs dedicated customer identities, permissions, write quotas, send idempotency, and inbox state. No current sending endpoints.
Managed reads are implemented and require only a ProfileGlide API key from the customer.
Boundary / next step: Bring-your-own-account connection, reauthentication, private data isolation, and identity billing are not implemented. One API key does not create identity permissions.
All 16 actions accept /run/live and /run/async with the same input schema; poll /v1/runs/{id}.
Boundary / next step: Live may return 202 while waiting for capacity. Read status and error, including on HTTP 200. Async does not add pagination.
Submit separate profile runs and collect their outputs in your application.
Boundary / next step: No array-input batch endpoint. Each profile is 20 credits; one active run and at most 20 unfinished runs per workspace. No 40 req/s promise.
The same idempotency key and body identify one logical run. Changed inputs under that key return 409.
Boundary / next step: This is request deduplication, not automatic provider retries or an SDK backoff engine. A terminal failed run stays terminal on replay.
The current response exposes completeness.has_more and a single_page scope.
Boundary / next step: No public continuation cursor, automatic pagination, or million-record job. Needs cursor binding, page-level reservations, deduplication, and cancellation tests.
Poll runs and /v1/runs/{id}/outputs to retrieve results today.
Boundary / next step: Webhooks need signed deliveries, durable retries, replay logs, and destination/SSRF controls. No callback URL parameter is accepted yet.
Your scheduler can submit new reads, persist snapshots, compare fields, and notify your team.
Boundary / next step: No hosted cron, built-in diff service, or change feed. Reuse a key for a retry; use a new key for a new observation.
Compose search → enrich → filter → store using your application or an external orchestrator.
Boundary / next step: No workflow designer, durable multi-step engine, or send/invite step. Source limits and per-action credit charges still apply.
REST + JSON, Bearer API-key authentication, cURL examples, and an OpenAPI 3.0.3 document.
Boundary / next step: No published TypeScript SDK, SDK-managed retries, or OpenAPI 3.1 claim. Use fetch/your HTTP client or validate a generated client against the spec.
The backend schedules approved sources, enforces leases and budgets, handles cooldowns, and keeps provider credentials server-side.
Boundary / next step: Availability depends on eligible source accounts. An exhausted quota or denied route does not justify bypassing controls through raw requests.
Verified Supabase Auth, revocable hashed API keys, tenant-scoped access, atomic credit reservations/refunds, and distributed request limits.
Boundary / next step: These implementation controls are not a SOC 2 certification or proof that public production infrastructure is deployed.
Bounded live cloud tests exist for the 16 data actions and accounting behavior.
Boundary / next step: No representative p50/p99 benchmark, uptime SLA, independent SOC 2 audit, or customer-selectable residency contract. These require operating evidence and external work.
Source reads return records available to the approved account and requested action.
Boundary / next step: We have not established these population sizes or universal visibility. Private, missing, and source-truncated data cannot be promised.
The backend grants 100 credits once per verified new workspace. Proposed Scale pricing tops out at 300,000 customer credits.
Boundary / next step: Public signup/deployment and paid checkout remain launch work. Shared source capacity is 300k records/month, not a per-customer guarantee or 300k full profiles.
For data reads, evaluate Sales Navigator search and parameter lookup first, then native Unipile reads, then a fixed, validated raw adapter only where neither supplies the required data. Raw access cannot add missing permissions, reveal anonymous users, or bypass account quotas.
Publish a modified equivalent under its actual meaning: search-visible members, search-derived distributions, or firmographic similarity. Do not rename these as exhaustive rosters, exact analytics, or ownership relationships.
Next priorities: structured people/company search, cursor pagination, and richer verified field projections; then signed callbacks and schedules. Connected-account outreach is a separate product expansion. A published SDK follows a stable contract. SOC 2 and latency/uptime guarantees require audits and operating evidence.
Basis: the supplied Edges screenshots, our current action contracts and output projections, and official Unipile search and company detail documentation. Provider documentation establishes candidates for validation, not current ProfileGlide support.
Read the eight action contracts →