Comparison
LibreChat is a polished self-hosted chat interface across many LLM providers. Pinchy is an agent platform built for teams. Here's how they actually differ — and when to pick each.
The Core Difference
A multi-provider chat UI. Users pick a model (OpenAI, Anthropic, Ollama, Google), start a conversation, attach files, enable plugins. Focused on the individual chat experience.
Agents are the primitive — roles with named knowledge, scoped plugins, and per-user access. Users don't pick a model; they pick an agent. The model is an implementation detail behind the role.
Many teams run LibreChat for exploratory chat and Pinchy for production agents. Both are open source, both are self-hosted, and both can talk to the same Ollama backend.
Side by Side
| LibreChat | Pinchy | |
|---|---|---|
| Primary primitive | Chat session + model | Agent (role + tools + scope) |
| Multi-provider support | Excellent (many providers) | OpenAI, Anthropic, Google, Ollama (local + cloud) |
| Self-hosted | Yes (Docker) | Yes (GHCR pre-built images) |
| Plugins / tools | Per conversation | Per agent, allow-listed |
| RAG / documents | Per conversation | Knowledge base per agent + group |
| Role-based access control | Basic user roles | Built-in (Enterprise) |
| Group-based agent visibility | Not the focus | Yes (Enterprise) |
| Audit trail | Conversation logs | HMAC-signed tool-call trail |
| Approval / human-in-loop | Manual | Allow-list + draft-only (e.g. draft without send); per-template confirm prompts |
| Usage & cost dashboard | Per-user | Tokens & cost per agent/source |
| External channels (Telegram) | No | Per-agent bots |
| Business-system integrations | Plugin API | Odoo first-class + plugin architecture |
| License | MIT | AGPL-3.0 open source |
Being Honest
LibreChat has broad provider coverage — OpenAI, Anthropic, Google, Mistral, Azure, Ollama, and more. If your core need is switching between cloud LLMs in one UI, it's hard to beat.
Conversation presets, prompt library, multi-modal support. It's one of the nicest open-source chat UIs available, and it's a pure MIT-licensed project.
If the goal is "give knowledge workers a better ChatGPT that I control", LibreChat hits that bullseye. Low configuration, immediate value.
Where Pinchy Wins
Each agent is a role (Quote Drafter, HR Onboarding, Compliance Checker) with its own tools, data, and users. Users don't configure a conversation — they pick the colleague.
Who sees the agent, which data it reads, which tools it calls, whether it sends or drafts — all first-class configuration. The permission layer enforces before the model ever runs.
Groups for Engineering, HR, Finance. Each agent is visible to the groups that should see it. Nobody stumbles into the HR agent from Engineering — it doesn't appear in their sidebar.
Tool calls, knowledge-base hits, user interactions — every event is signed with a per-row HMAC-SHA256. Append-only, verifiable row-by-row. When an agent takes action on your business, the trail shows exactly what it did and why.
Not "upload docs to your conversation". Knowledge is attached to the agent, scoped by group, and retrieved with citations. HR docs stay with the HR agent.
First-class Odoo integration. Plugin architecture for connecting your own software. Built for agents that act on the business, not just chat about it.
Decision Guide
You want a polished multi-provider chat UI for individuals. The goal is better personal productivity with LLMs. Permissions and audit trails aren't the priority.
You're deploying AI agents for a team. Different users need different agents. Agents act on business systems. Compliance asks for a trail. Boundaries matter.
Individuals want LibreChat for exploratory chat; the company needs Pinchy for production agents. Shared Ollama backend, separate concerns.
FAQ
They target different goals. LibreChat is a self-hosted, multi-model chat interface — ChatGPT-like UI with plugin support and multi-provider routing. Pinchy is an agent platform: roles, scoped permissions, knowledge bases, audit trail, and per-agent channels. If your need is 'a great chat UI for multiple LLMs', LibreChat is strong. If your need is 'AI agents with boundaries for a team', Pinchy is the right shape.
LibreChat has user authentication and basic role features, but permissions are mostly at the account level, not at the agent-with-tools level. Pinchy is built around agents as the primitive — each agent has its own tool allow-list, knowledge scope, and group access control. Different problems the two projects prioritise.
Yes. Both can point to the same Ollama instance. Teams often run LibreChat for individual model exploration and Pinchy for team-facing agents, both hitting the same local LLM infrastructure.
LibreChat logs conversations and usage per user. Pinchy adds tamper-evident, HMAC-SHA256-signed audit entries for every tool call, every knowledge-base hit, and user interactions — an append-only trail you can verify row by row. Pinchy doesn't ship a built-in approval workflow; control comes from each agent's tool allow-list (zero tools by default) plus draft-only conventions, so an agent can be scoped to draft without sending. That's the kind of trail compliance teams need when agents act on behalf of the business, not just chat with users.
Self-host Pinchy yourself in minutes, or book a call to talk it through. Your choice.
Or email us: info@heypinchy.com