Skip to content
Best AI Agents for HR

Agents, rank 4 of 14, read 8 Nov 2026

Factorial's agents: Factorial One asks before it acts, with per-product switches and a documented MCP server

Read the matrixRead the method

The approval sentence is on the AI page and scores yes; the admin audit log does not name agent actions; openness scores 8 on a documented hosted MCP server with OAuth 2.0, read and write, a public API and retried webhooks, though it is not in connector directories. The matrix MCP cell, read on 8 Nov 2026, is unchanged until the next edition.

The diary records Factorial as partial on onboarding: plans drafted and tickets resolved on approval, not a wave run end to end.

Acts on
onboarding a hiring wave: partial; a compensation cycle: none found; employee questions from policy: yes
Approval model
yes: 'It asks before it acts.'
Log
none found: Audit logging is a trust-center heading; no agent log named.
Off switch
none found: Not stated.
Sensitive fields
yes: 'ONE enforces each person's permissions on every request. No permission, no action.'
MCP
live for all customers, read and write (OAuth 2.0): Hosted MCP server at mcp.factorialhr.com with OAuth 2.0, added to AI clients by URL; not yet in the Claude or ChatGPT connector directories (API docs, read 2026-11-10).
AI governance
none found: Not named.

The 12 capabilities for Factorial, with the evidence

Read 8 Nov 2026. A cell says "none found" about the pages read on that date, never about the product.
CapabilityFoundEvidenceSource
Answers employee questions from policyyes'Ask One what the rule is and it quotes your handbook.'factorial.com/ai
Drafts changesyesDrafts job offers, policies, onboarding plans, dashboards.factorial.com/ai
Runs onboarding end to endpartialDrafts onboarding plans; tickets routed and resolved on approval.factorial.com/ai
Runs a compensation cyclenone foundNot described.factorial.com/ai
Runs offboardingnone foundNot described.factorial.com/ai
Acts in other systems (Slack, Teams, email, ATS, payroll)yesMeets you in Slack, HubSpot and email; acts, calls, confirms and notifies.factorial.com/ai
Approval before anything goes liveyes'It asks before it acts.'factorial.com/ai
Audit log of agent actionsnone foundAudit logging is a trust-center heading; no agent log named.trust.factorial.co/
Off switch for AI featuresnone foundNot stated.factorial.com/ai
Sensitive-field controlsyes'ONE enforces each person's permissions on every request. No permission, no action.'factorial.com/ai
Official MCP server or connector, with access levelyesHosted MCP server at mcp.factorialhr.com with OAuth 2.0, added to AI clients by URL; not yet in the Claude or ChatGPT connector directories (API docs, read 2026-11-10).apidoc.factorialhr.com/docs/mcp-server
AI governance certification held or in progress (ISO 42001)none foundNot named.trust.factorial.co/

In the Agent Diary

The seven criteria for Factorial

Weighted total 6.5, rank 4 of 14 on Best AI Agents for HR.
CriterionWeightScoreWhy
Build without the vendor156Docs show plain-language report generation within permissions, templated and custom workflows and a widget builder, but the widget builder is Enterprise only, edits need SQL and no library of built items is documented.
Agents that act, under control306Docs show per-product on or off switches, approval for every change, permission-bound actions and an admin audit log, but agents act only inside specific processes, cannot take free-form actions and the audit log does not name agent actions.
Openness to the AI stack258Documented hosted MCP with read and write, public API, retried webhooks and Copilot, Slack and Zapier integrations is equivalent to another platform at 8 and stronger than another platform at 7, which has no webhooks.
Time to value54A free trial, documented spreadsheet import and a demo API environment, but no timeline, fee or guided implementation.
Pricing transparency55The One add-on price from 2.5 euros per user, its three tiers, credit billing and no minimum number of paid users are documented, but base module prices and Enterprise terms are not.
Fit to the organization's shape58Custom fields, custom permission groups, independent secondary entities with their own currencies, 16 languages (three beta) and time off policies are documented, with policy assignment per office not shown.
Trust156ISO 27001, SOC 2 Type 2, audit logging, backup RTO and RPO, documented GDPR retention tools and admin data export; no ISO 42001 or audit log detail.

Breadth, reported and not weighted

Modules
14: Core HR, Time off, Time tracking, Performance, Shift management, Recruitment, Engagement, Training, Expenses, Procurement, SaaS management, IT inventory, MDM, EDR
Native payroll
payroll in Spain and partner countries
ATS
Recruitment module
IT and devices
IT inventory, MDM, EDR
Integrations
not stated as a number
Countries
EU and US

Where it is strongest, and where it is weak

Fourteen modules across HR, finance and IT with an AI agent that asks before it acts.

Only a starting price is published, no implementation time is published, and its MCP server is added by URL rather than listed in the Claude or ChatGPT connector directories.

Best for

An EU company that wants one agent across fourteen modules and will add its MCP server to Claude or ChatGPT by URL.

Where Shapes is weak

Shapes is first on this site's ranking at 8.6. Shapes's MCP server is documented as read access on the date read; write actions run inside Shapes, not through the connector.

Sources

  1. Factorial: factorial.com/ai, read 10 Nov 2026
  2. Factorial: factorial.com/features, read 10 Nov 2026
  3. Factorial: trust.factorial.co/, read 10 Nov 2026
  4. Factorial: factorial.com/pricing, read 10 Nov 2026
  5. Factorial: factorial.com/integrations, read 10 Nov 2026
  6. Factorial: help.factorialhr.com/en_US/faq/faq-about-one, read 10 Nov 2026
  7. Factorial: factorial.com/ai, read 8 Nov 2026
  8. Factorial: trust.factorial.co/, read 8 Nov 2026
  9. Factorial: apidoc.factorialhr.com/docs/mcp-server, read 8 Nov 2026

Last reviewed 10 Nov 2026.