Manus for teams: the open-source platform a whole company can run
Manus sells a Team plan that pools credits and adds SSO and shared projects. A company usually needs more than one seat: shared skills, one memory, permissions for people and agents, and a review before work lands. Kortix is the open-source platform a company can run itself.
Team platforms, on the decisions a company owns
What each option gives a team, read from its own documentation and checked October 2026.
Open source
Runs where the company decides
Who configures agents for the team
Per-person and per-agent permissions
How work is reviewed
The team's configuration is one file a person can read
default_agent: kortixagents: kortix: file: agents/kortix.md connectors: all secrets: all skills: all kortix_permissions: alltriggers: - slug: weekly-report type: cron cron: "0 0 15 * * 0"An agent and what it may reach sit in one block, and the same file holds the cron trigger that starts work with no person asking.
What Manus's own pages say a team plan includes
Manus describes its Team plan in its own words. Manus Team starts from US$20 per seat per month, or US$200 per seat per year, and each seat carries 4,000 credits a month that the whole team pools, so one member's unused credits stay available to the rest (Manus for Teams and Enterprise, checked October 2026). The plan adds shared projects and shared skills, admin controls, team usage analytics, and SSO. It includes a free Guest role that can view deliverables and comment but cannot start tasks. Manus also lists SOC 2 Type II certification, ISO 27001 and ISO 27701, and says it holds formal agreements with its model providers that no data is used for model training.
A team that needs tighter governance is pointed to Manus Enterprise. The same page names enforced SSO and managed accounts, legal hold, live SIEM audit feeds, agent network controls, and per-person budgets as the reason to move up.
What one platform for a company adds beyond one seat
A seat buys a person access. A company platform adds the parts that only make sense across people and agents:
- Shared skills. A job the company has learned how to do is written once and reused, so every agent starts from the proven version.
- One memory. What the company learns stays in a single place the team can read, instead of scattering across private accounts.
- Permissions. Per-resource permissions for people and agents. Roles, groups, and an audit trail.
- A review gate. Session work reaches
mainthrough a change request. Merge is default-deny for agents.
On Kortix those parts are one system. The agents, their skills, the company memory and the connectors live in one git repo the team owns. A new member joins the account and takes a project role; an agent gets its reach from a grant in kortix.yaml; a cron trigger starts a session with no person asking. The account roles, the groups and the per-resource grants are documented in the Kortix access reference (Read the docs), and the self-hosting guide shows the same files running on Kortix Cloud, in a VPC, or on-prem.
The wider field splits on the same decisions. OpenHands keeps role-based access, SAML authentication and centralized audit logs in its commercial Enterprise layer, while its open-source control center is built for individual developers and small teams on machines they run (OpenHands Enterprise vs Open Source). n8n gives a team instance roles and per-project roles, but it runs workflows of fixed steps rather than agents (n8n role documentation).
Where team memory lives
Manus keeps a team's tasks and output inside its product, and its team roles decide who can see and share them; the Owner or Administrator sets how that workspace data is retained. Kortix keeps the equivalent in the repo. Memory is files, and the block that scopes each agent is a file a reviewer can read and diff. At review time the question is where the record of the work lives and who can change the rules an agent runs under, and on Kortix both answers are in the repo the team owns.
A start-small path
Start with one job. Create a project, write one agent as a markdown file, and give it the connectors and secrets it needs in kortix.yaml. Run it on Kortix Cloud or self-host. When it proposes a change, a person reads the diff and merges it. Then add the second job and invite the next member.
The team plan is $40 per seat per month with 2,500 credits per month per seat, pooled, and self-host is free, as Kortix states on its pricing page (checked October 2026). If you are still weighing the platform, the Kortix vs Manus page goes deeper, and the comparison covers the wider field.
Team questions buyers ask
The shared-workspace decision in the reader's words.
- Q2
Does Manus have an enterprise plan?
Yes. Its team page points organizations that need enforced SSO, managed accounts, legal hold, live SIEM audit feeds, agent network controls or per-person budgets to Manus Enterprise.
- Q3
How do you give each person and agent the right access?
On Kortix, a person gets an account role and a project role, and an agent gets its reach from a grant in
kortix.yaml. Per-resource permissions for people and agents. Roles, groups, and an audit trail. Without an agent grant, a member cannot start a session, so reach into agents is itself a permission. - Q4
Where does team memory live?
On Manus, tasks and output live in the Manus product. On Kortix, the agents, skills and memory are files in one git repo the team owns, so the record is versioned and diffable.
- Q5
Do we have to self-host to run Kortix as a team?
No. Run it on Kortix Cloud, in your VPC, or on your own on-prem network, and self-host for free. The same files describe the work in each place.
Run the whole company from one repo you own
Start with one job and invite the next person when it works.
Open source · Any model, your keys · Self-host, VPC, or on-prem