Manatal

Named outside systems receive writes across six classes, from assessments and e-signature to the customer's own mailbox. For HRIS and payroll the list includes BambooHR, HiBob, Humaans, SAP SuccessFactors, Oracle Taleo and ADP, with SAP, Oracle and Kronos named again in the pricing FAQ. Job boards and sourcing run through LinkedIn and LinkedIn Recruiter, Indeed, Monster, ZipRecruiter, SEEK, Xing and GitHub, with distribution to "thousands of free and premium channels."

Outlook, Office 365 and Google Workspace connect both ways for sending, receiving and scheduling. Codility handles assessments and Adobe Sign handles signatures, while Mailchimp and other providers cover marketing. Social posting reaches Facebook, LinkedIn, WhatsApp, WeChat and Line, and Zapier connects 3,000+ apps while n8n reaches Slack, Salesforce and Microsoft Teams.

The writes are documented, not just the catalog. Assessments are "sent to candidates in one click with test results received back into Manatal," e-signatures are requested and signed documents filed onto profiles automatically, and emails and SMS go out through the customer's own connected mailbox and number.

The public API documents create, update and delete operations on candidates, jobs, matches, notes, attachments and organizations, and the MCP server writes as well as reads, creating notes and managing matches. These are authenticated writes into systems Manatal does not own and records the customer owns.

Zapier, n8n and Open API access are Enterprise Plus only, so the two broadest routes are missing on Professional and Enterprise; native connectors, mail and calendar, and job boards and SMS come with every tier. No page confirms the figure of over 2,500 job boards.

Sourcemanatal.com/features/applicant-tracking-system, /pricing and developers.manatal.com/referenceread 2026-09-12

Customers author the pipeline and automate steps within it, but recruiters drive it, and nothing plans work or hands it between the AI units. Pipelines are unlimited and let you "fully customize your recruitment stages for each job." Workflow automations "trigger at any step of the hiring process" and can be set across all jobs or per job, and screening and knockout questions gate entry.

One step runs unattended from start to finish: the AI Interviewer conducts a full screening interview, scores the responses against the job's requirements and returns an assessment with no recruiter present. The participants are distinct: the recruiter who owns the pipeline, the hiring manager on a shared seat with scorecards, and the candidate, with the AI units acting on records.

What is missing is planning. Manatal ships seven named AI units but documents no handoff between them, no orchestrator and no plan built for each request. The parser, scorer and Interviewer each act on their own trigger, as do the enricher and Copilot, and the recruiter moves a candidate from one to the next by dragging a card across a stage. The automations deliver on rules when state changes; no agent decides what to do next. Workflow automations come with Enterprise and above, so the Professional plan has customizable stages and no triggers at all.

Sourcemanatal.com/features/applicant-tracking-system, /features/manatal-ai and /pricingread 2026-09-12

The agents score, search and recommend over the customer's own candidate database, which the product keeps as a structured, indexed store. Its "scoring and matching AI agents extract the requirements from your job description, then score every candidate in your database against them."

AI Advanced Search combines traditional filters with "intelligent semantic search" that "goes deeper than keywords," returning relevance scores, with Boolean search alongside. AI Recommendations "proactively recommend gems from your talent pool" and suggest jobs back to candidates, with weighting and criteria the customer can adjust.

The store is maintained, not assembled for each run. Resumes are parsed into structured profiles the moment they are uploaded, skills, experience and education from CVs and LinkedIn are indexed on the profile, and enrichment writes public professional data back in. The ATS page says the database keeps a well organized record of applicant profiles "accessible months or years later."

It scales far past any context window across an enterprise's whole candidate estate and stays searchable, though it is not a retrieval engine built on citations. A separate external sourcing hub of 700M profiles is Manatal's own index for finding new candidates, and the MCP server also lets assistants query in natural language.

Sourcemanatal.com/features/manatal-ai and /features/applicant-tracking-systemread 2026-09-12

The AI produces scores, recommendations and interview assessments for recruiters to review, but nothing holds the agent's own actions for approval. Evaluations from the AI arrive in Manatal for review, candidate quick screen lets a recruiter preview and assess candidates and then "advance them to the next stage or drop them," and the AI Interviewer's assessments land on the profile.

Each of those is a person acting after the agent has finished, and scorecards standardize how the team evaluates. Configuration before the run does not gate anything either: the interview builder, screening and knockout questions, and weighting and criteria shape what the agent will do, which is setup rather than oversight.

There is no approval queue, no checkpoint the agent stops at and no threshold that escalates on confidence or risk. Nothing documented requires sign off before the AI Interviewer invites a candidate and runs a screening interview unattended 24/7, which is the agent's own consequential action. Manatal documents review of the agent's output, not a stop on its actions. Role based permissions govern access rather than actions.

Sourcemanatal.com/features/manatal-ai, /features/applicant-tracking-system and /pricingread 2026-09-12

Certification is SOC 2 Type II, role based access control comes with every tier, and single sign on is kept for Enterprise Plus. The security page states "Manatal is SOC 2 Type II certified, verified by a top third-party auditor," and the footer of every page carries the AICPA SOC 2 badge alongside GDPR and CCPA badges.

Access control reaches every tier, not just the top one: "all data access is protected by a role-based access-control mechanism, which only lets users view data for which they have permission." The plan matrix lists layered user permissions and roles, two factor authentication and auto log out policies that admins enforce, on Professional, Enterprise and Enterprise Plus alike.

Single sign on, customizable user permissions and user group management are Enterprise Plus only, so a buyer on Professional or Enterprise gets RBAC and 2FA but no identity provider integration. Manatal also publishes a vulnerability disclosure program, SSL in transit and hashed passwords, along with white box security assessments, Stripe as its payment subprocessor at PCI Level 1, and tested backups kept 30 days.

Sourcemanatal.com/security and /pricing plan comparisonread 2026-09-12

Access to customer data is logged and audited, and an activity trail can be retrieved, but nothing records what the AI units did. Under Application Monitoring the security page states that "all access to Manatal is logged and audited," and adds something the customer can ask for: "you can ask us at any given time for a report of who accessed your data, when and why." Activities Management covers calls, emails and meetings, along with interviews and tasks, with a calendar and reminders.

The public API treats activity as its own object model with endpoints per candidate, job, match, contact and organization, so the trail can be pulled by program, not just seen on screen. The AI Interviewer's assessment stays on the candidate profile, the AI Copilot's fit and gap analysis shows how a candidate matches, and reporting runs deep across dashboards, KPIs and an advanced custom report builder on Enterprise Plus.

None of it is scoped to the agent.

Nothing says that the AI units' actions, such as which candidates the scoring agent ranked and on what weighting, when the enrichment agent wrote to a profile and from which source, which automation fired and why, or what the Interviewer asked and how it reached its score, are captured as a record that can be reconstructed and attributed to the agent rather than a user.

Access logs answer who opened a record and the activity trail answers what happened to it; neither says what the agent did and why. The reporting suite measures the hiring funnel, which is the customer's own operations.

Sourcemanatal.com/security, /features/applicant-tracking-system and developers.manatal.com/reference/activityread 2026-09-12

There is a customer data store but no agent memory. The ATS keeps full candidate profiles covering resume details, messages, annotations and vacancy feedback, "retains past data for analysis," and keeps resume records "accessible months or years later." That is the customer's application data, the thing the product exists to hold, and it is the corpus the scoring and search agents read.

No agent context survives between runs, and no preference set carries across them. There are no decision traces and no stated scope or lifetime for anything the AI holds, and the customer cannot inspect, edit, export or purge agent state apart from the candidate records.

The AI Copilot is called "context-aware," which describes awareness inside a conversation and says nothing about persistence beyond it. Each AI unit acts on the current job, resume or profile: the scorer pulls requirements from the job description and scores the database, the parser reads the file in front of it, the enricher updates a profile and the Interviewer runs one interview. Nothing builds up across them.

Sourcemanatal.com/features/applicant-tracking-system and /features/manatal-airead 2026-09-12

All customer data is stored in the USA, with no way to choose another location. The security page says so under Customer Data: "All customer data is stored in the USA." No second region is named anywhere. The plan comparison table itemizes three tiers line by line and has no region, residency, data location or hosting row at any of them, not even on Enterprise Plus, where SSO, user groups, API access and the advanced report builder are the listed differentiators. No page offers an environment the customer controls, an EU option or any way to choose.

The infrastructure is Manatal's alone: all services run on Amazon Web Services, in a tiered network Manatal designed, with daily backups kept for 30 days. The pricing FAQ answers "Is Manatal available in my country?" with a cloud platform "available globally, without restrictions of geography," which describes where the product can be reached from, not where data sits; the security page settles that. Manatal sells GDPR and PDPA compliance tooling into 135 countries from storage in the US only.

Sourcemanatal.com/security and /pricingread 2026-09-12

Seven distinct named AI units ship, each a separate agent a buyer can take or leave. The AI page gives each its own section and knowledge base entry. The AI Interviewer conducts screening interviews unattended in six named languages, the AI Notetaker joins Google Calendar and Microsoft Teams calls to record, transcribe and summarize, and the AI Copilot offers context aware chat, candidate summaries, fit and gap analysis, an interview prep builder, and compare and rank.

AI Candidate Scoring and Matching and AI Resume Parsing are units of their own, and so are AI Profile Enrichment and the AI Job Description Generator. Take away the Notetaker and the Interviewer still works whole, with its own configuration and output; take away the Interviewer and the Copilot, scorer and parser are untouched. These are separate units doing separate jobs, not modes of one engine.

A buyer selects among them rather than getting one block. The plan matrix carries the AI Interviewer as its own add on section and the MCP server as an Enterprise Plus toggle, and the People-Match extension is a separate browser install.

Template banks come with every tier, among them job templates, shareable email templates and career page templates with advanced branding, alongside a custom resume builder and candidate scorecards. No single catalog page lists the units for browsing; they are spread across the AI feature page and the plan comparison table. Each unit's own entry point may sit in the knowledge base at support.manatal.com.

Sourcemanatal.com/features/manatal-ai and /pricing plan comparisonread 2026-09-12

Workflow automations fire on pipeline events, the AI Interviewer runs around the clock and candidates are reached across many channels, though automations start at the Enterprise tier. Events are configurable: "create workflow automations to trigger at any step of the hiring process," sending personalized messages "as candidates progress through stages, or are dropped from your pipeline," and automations can be "configured to run across all jobs or customized for specific jobs."

Career page applications write straight into the pipeline with all fields synced, forwarded resumes create candidates, and the AI Notetaker joins scheduled Google Calendar and Microsoft Teams calls when the calendar event comes up. The AI Interviewer conducts "24/7 automated video screening" with no one starting each run, and activity reminders and a weekly account summary run on their own schedule.

The channels are many and named. Text messaging and SMS come on every tier, email runs through connected Outlook, Office 365 or Gmail with mass mailing and marketing tool sync, and the branded career page sits on a custom domain.

Candidates and employees have their own portals, social posts reach Facebook, LinkedIn, WhatsApp, WeChat and Line, and job distribution reaches "thousands of free and premium channels." Video interviews, notifications in the app and by email, and a progressive mobile app round it out.

Workflow automations are Enterprise and above, so on Professional the channels exist but the trigger framework does not, and the cheapest published plan buys manual sending.

Sourcemanatal.com/features/applicant-tracking-system, /features/manatal-ai and /pricingread 2026-09-12

No model is named behind the agents, and the customer gets no choice of one. The MCP server lets recruiters connect ChatGPT, Claude, Gemini or Microsoft Copilot to Manatal, but that runs the other way: the customer's own assistant calls Manatal to read and write recruitment data, and it does not let anyone choose which model runs Manatal's AI.

There is no model picker, admin setting or bring your own key path, no model assigned per agent, and no disclosed routing on the vendor's side. No provider is named for the AI Interviewer, the scoring and matching agents or the Copilot, nor for the parser, enricher or job description generator.

The plan comparison table itemizes the platform line by line across three tiers, down to auto log out policies, mailbox integration, duplicate management and interface language, and its only model row is "ChatGPT, Claude and LLM integration," which is the MCP row. Over the models running the agents, the customer chooses nothing and is told nothing.

Sourcemanatal.com/pricing, /features/manatal-ai and /securityread 2026-09-12

A complete public API reference covers the platform, along with webhooks and an MCP server that writes as well as reads. The developer site at developers.manatal.com is linked from the footer and the Resources menu of every page. It documents an API rooted at api.manatal.com/open/v3, with general sections on authorization, rate limiting, webhooks, pagination, link validity, error handling, formats, custom fields and file upload, and on testing with Postman.

Thirteen object models each list their read, create, update and delete operations one by one, covering candidates, educations, experiences, activities, contacts, attachments, jobs, notes, matches, users, organizations, application forms and job pipeline stages. A separate career page API covers job posts, application forms and applying or referring, and a webhooks service called Manahook ships its own create, update and delete endpoints, plus verify and get, so the platform pushes as well as receives.

The MCP server searches candidates and jobs, creates notes and manages matches for any MCP compliant client at 100 requests per minute, and Zapier and n8n connect too. An llms.txt is published at developers.manatal.com/llms.txt. Three limits apply. Open API access and the MCP server are Enterprise Plus only on the plan matrix. V3 endpoints need a token from the support team, a credential step rather than a documentation gate. And the reference was last updated about a year ago and calls itself a work in progress covering a subset of the platform.

Sourcedevelopers.manatal.com/reference/getting-started, support.manatal.com/docs/mcp-server and manatal.com/pricingread 2026-09-12

No way to test or measure the agents is documented; what gets measured is candidates and the hiring funnel. Predictive analytics and candidate scoring make screening more systematic and data driven, but candidate scoring is the agent's output about candidates, not a measure of the agent.

The reporting suite measures the customer's hiring funnel, covering time to hire, cost of hire, rejection reasons, acquisition by channel and source, hiring performance ratios, leaderboards and recruiter performance, which is the customer's own operations. Candidate scorecards are a form for people to compare candidates, not a harness.

The nearest control is "customizable weighting and criteria to fine-tune candidate recommendations" on every tier, which configures the agent but reads nothing back: no feedback signal, no measured result, no before and after comparison.

No test environment, sandbox, holdout set or versioning appears anywhere, nor a regression check, evaluator model or scored result about agent behavior. The AI Interviewer screens candidates unattended and the scoring agent ranks an entire database, with nothing measuring either. The knowledge base at support.manatal.com or the AI Interviewer feature page may carry a way to calibrate assessments.

Sourcemanatal.com/pricing, /features/manatal-ai and /features/applicant-tracking-systemread 2026-09-12

The only browser piece is an extension that captures profiles the recruiter is already viewing, and no browser or computer use is documented. The People-Match AI browser extension lets recruiters "source and import candidates from LinkedIn and 10+ other platforms, directly into Manatal as you browse," with "one-click profile creation without ever leaving the candidate's page." The person decides which profile is open and the extension reads it into Manatal, which is ingestion through a browser, not an agent driving an interface.

The AI Interviewer runs video interviews 24/7, but that session is Manatal's own surface, not an interface it drives for the customer, so changes to someone else's UI cannot break it. No hosted or local browser, desktop session or computer control appears, and there is no RPA fallback. Manatal reaches other systems through code, over 200 documented endpoints and webhooks, plus Zapier and n8n, none of which navigates anything.

Sourcemanatal.com/features/manatal-ai, /features/applicant-tracking-system and /pricingread 2026-09-12