Feature
Every agent has a name, a character, and clear boundaries.
Configure personalities, knowledge bases, and permissions from one settings panel.
Every agent is configured through a single settings page. No config files, no CLI, no YAML. Click the gear icon and everything is right there.
Name, tagline, and AI model. Pick from any configured provider — OpenAI, Anthropic, Ollama, or your own. The model dropdown groups by provider so you always know what you're choosing.
Choose a preset or write a custom SOUL.md. The personality defines how the agent speaks, what tone it uses, and what character it has. Smithers is dry and competent. Your HR agent can be warm and encouraging. Your DevOps agent can be terse and precise.
AGENTS.md defines the agent's operating instructions — what it should do, how it should approach tasks, what resources it has. Think of it as the job description.
Allow-list of tools the agent can use and directories it can access. Nothing is enabled by default. You grant exactly what's needed, nothing more. Learn more →
Who can talk to this agent? Everyone, specific groups, or nobody yet. Enterprise feature that maps agent access to your org chart. Learn more →
Generic chatbots feel like tools. Named agents with consistent characters feel like colleagues. That's intentional.
Pinchy's default agent — Smithers — is competent, polite, and occasionally dry. He greets new users with an onboarding interview to understand the company and set up the platform.
But Smithers is just the starting point. Every agent you create gets its own personality:
Clear, efficient, focused. Gets to the point. Perfect for executive-facing agents that summarize data and draft communications.
Precise, detailed, opinionated about best practices. Ideal for code review agents, DevOps helpers, and architecture advisors.
Warm, patient, encouraging. Great for onboarding agents, HR assistants, and any agent that works with non-technical teams.
Write your own SOUL.md from scratch. Define tone, vocabulary, boundaries, quirks — whatever makes the agent feel right for your team.
An agent without context is just a general-purpose chatbot. An agent with your company's documentation becomes genuinely useful.
Pinchy lets you mount specific directories into an agent's scope. The agent can read files in those directories — and nothing else. No accidental access to HR data from the engineering agent.
Select exactly which folders under /data/ the agent can access. A visual directory picker — no path guessing.
Agents read your documents but can't modify them. If you need write access, you enable it explicitly per tool.
Markdown, PDF, plain text, code files — agents read whatever you give them. Mount your wiki, your runbooks, your policies.
Created by admins. Visible to the whole team (or specific groups). Memory and conversations are shared — when one person teaches the agent something, everyone benefits.
Use for: Company knowledge bases, team assistants, department-specific tools.
Auto-created for each user. Private memory, private conversations. Your personal AI assistant that knows your context without sharing it with others.
Use for: Personal productivity, drafting, brainstorming, individual workflows.
30-minute demo. We'll walk through creating an agent, configuring its personality, and connecting it to your team's knowledge.