AnythingLLM limitations, and the open-source system a company needs
AnythingLLM's hosted cloud ships no built-in model, no custom agents and no MCP, and the open-source system a company turns to instead is Kortix. AnythingLLM's limits come from its own documentation, split by how you run the app.
In the repo: agents, triggers and connectors as files
# kortix.yamlagents: support-triage: file: agents/support-triage.md # what it does, in plain markdown connectors: all # what it may reachtriggers: - slug: daily-digest type: cron cron: '0 0 9 * * 1-5'A Kortix project keeps its agents, triggers and connector configuration in one kortix.yaml, beside the code the company already owns.
What AnythingLLM's own documentation says about its limits
AnythingLLM is a document-chat app from Mintplex Labs. You point it at your files and talk to them, with agents and retrieval built in. It ships in three forms: a desktop app that runs on your own machine, a self-hosted server you run with Docker, and a managed hosted cloud on a custom domain. Its own documentation is specific about where those forms differ.
The hosted cloud carries named limits, listed on AnythingLLM's own cloud limitations page:
- It ships no built-in model you can use, so you connect your own local or cloud provider.
- The built-in embedder can overwhelm the instance on a large document and return an error.
- The "Accuracy Optimized" search can leave an instance slow or unresponsive when one knowledge base holds many files.
- It supports no custom agents.
- It supports no MCP.
Multi-user support and permissioning are part of the Docker version, and the desktop app stays single-user, as AnythingLLM's own repository states. Custom agents and MCP are available on the desktop and self-hosted forms, and turned off in the hosted cloud. The hosted cloud is designed for the quickest start, and it carries the shortest feature list of the three forms.
Where a document-chat app stops for a company
A document-chat app answers questions over the files you give it. The layer around that chat is what a team needs next: agents that act inside its tools, one memory the whole company shares, permissions for people and agents, and a review step before an agent changes a live system.
Kortix is the open-source AI Operating System: your agents, their skills, your company memory, and every connector in one git repo you own, with the agents working on real cloud computers. Where a document-chat app keeps its state inside the product, Kortix keeps it in files you can read and diff. Per-resource permissions for people and agents. Roles, groups, and an audit trail. Connectors reach 3,000+ apps in a click, plus MCP, OpenAPI, Postman, GraphQL and raw HTTP, and connector credentials are brokered server-side and never enter the machine. One isolated sandbox per session. Each session has its own isolated machine and branch. Session work reaches main through a change request. Merge is default-deny for agents.
The part a team needs once the work has a consequence is the part a document-chat app does not cover: taking actions in your tools and recording what was done. If you are weighing the two, the Kortix vs AnythingLLM comparison covers the same ground, and AnythingLLM for teams goes deeper on the team case.
What you get when the system is yours
Kortix runs any model provider with your own keys. Or the ChatGPT plan you already pay for. Or sign in with your OpenCode Console account for OpenCode Zen and Go. Run it on Kortix Cloud, in your VPC, or on your own on-prem network. Self-host is free. The whole configuration is plain files, as Kortix on GitHub shows, so you can read, fork and audit it.
If self-hosting is your path, self-hosting an AnythingLLM alternative covers the setup, and migrating from AnythingLLM covers moving your documents and agents across. For the wider field, a guide to self-hosted open-source AI workspaces compares several options. Read the guide.
A document-chat app and a company system, side by side
The Kortix column quotes Kortix; the AnythingLLM column comes from AnythingLLM's own documentation.
agents/<name>.md plus an agents.<name> block in kortix.yaml.AnythingLLM cells come from its own documentation and repository, checked October 2026. This site is not affiliated with AnythingLLM.
Questions people ask when AnythingLLM runs out of room
It ships no built-in model, supports no custom agents and supports no MCP. Its own documentation also says the built-in embedder, the "Accuracy Optimized" search and a large file count can overwhelm the small instance it provides.
Multi-user support and permissioning are part of the Docker version, and the desktop app stays single-user. The hosted cloud is multi-user and carries the same restrictions. A team that needs one shared memory, per-session isolation and a review step runs a system it owns.
Kortix is the open-source AI Operating System, so agents, skills, company memory, connectors and triggers live in one git repo you own. Per-resource permissions for people and agents, roles, groups, and an audit trail. Work lands through a change request a person reviews.
Yes. Your project is a git repo, so documents and company memory sit in files you can read and diff. The migration guide covers the move.
Run your whole company from one repo you own.
Start with one job and grow from there.
Open source · Any model, your keys · Self-host, VPC, or on-prem