What's the Difference Between /chat and a Custom Agent?
Last updated: April 7, 2026
/chat is your personal general-purpose assistant that automatically discovers every integration you've connected. A Custom Agent is one you build yourself — with a specific system prompt, a hand-picked set of tools, and the ability to be shared, deployed to Slack, triggered by events, and accessed via API.
The Short Version
Both /chat and Custom Agents are AI agents powered by the same underlying engine. Think of /chat as your ready-to-go personal assistant and a Custom Agent as a purpose-built specialist you design.
Capability | /chat (General Agent) | Custom Agent |
Custom system prompt | No — uses a built-in prompt | Yes — fully configurable |
Tools & integrations | Auto-discovers all your connected integrations here. | You pick exactly which tools it has |
Workflows as tools | No | Yes — attach any saved workflow |
Self-modification | No | Yes — can edit its own instructions |
Share with others | No — it's personal to you | Yes — share with anyone via the Share button (as Editor, Viewer, or Use Only) |
Deploy to Slack | No | Yes |
Event-based triggers | Can create triggers in conversation | Triggers configurable in agent settings |
API access | No | Yes |
Web search | Yes | Yes (when enabled) |
Image generation | Yes | Yes (when enabled) |
Code sandbox | Yes | Yes (when enabled) |
File uploads | Yes | Yes |
Skills | Loads all your personal skills | Attach specific skills |
When to Use /chat
Use /chat when you want a quick, personal assistant that can reach all your connected apps without any setup. Open it, ask a question, and it figures out which tools to use.
Quick one-off tasks: "Summarize my last 5 emails" or "What's on my calendar tomorrow?"
Exploring what Gumloop can do with your connected integrations
Ad-hoc work where you don't need a repeatable agent
/chat automatically sees every integration you've connected in Profile → Credentials — no configuration needed.
When to Build a Custom Agent
Build a Custom Agent when you need any of the following:
A specific personality or instructions — Custom system prompts let you control exactly how the agent behaves, what tone it uses, and what rules it follows.
A curated toolset — Instead of giving the agent access to everything, you pick only the tools it needs. This makes it more focused and less likely to take unexpected actions.
Workflows as tools — Attach saved workflows so the agent can run them as part of a conversation. The General Agent cannot do this.
Sharing or deployment — Share the agent with teammates, deploy it to a Slack channel, or embed it for external users.
Self-modification — Let the agent update its own instructions over time as it learns your preferences.
Model selection — Choose a specific AI model instead of the platform default.
To create one, go to Agents and click New Agent.
How Tools Work Differently
This is the biggest practical difference between the two:
/chat — Automatic Discovery
The General Agent uses dynamic tool discovery. When you start a conversation, it scans your connected integrations (Gumloop built-in integrations, Gumstack servers, and custom MCP servers) and makes all of them available. You don't configure anything — if you've connected Google Sheets, it can use Google Sheets.
Custom Agent — Explicit Configuration
A Custom Agent only has the tools you explicitly add in its Tools panel. This includes:
Gumloop integrations — Built-in integrations like Google Sheets, Slack, Gmail, etc.
Custom MCP servers — Your own or third-party MCP servers
Gumstack servers — Servers from the Gumloop marketplace
Workflows — Saved workflows the agent can execute
Built-in tools — Web search, image generation, and the code sandbox
This explicit approach is better for production agents because you control exactly what the agent can and can't do.
Skills Behavior
Both agent types support skills, but they discover them differently:
/chat automatically loads all your personal and project-level skills.
Custom Agents use only the skills you attach to them in the agent's settings.
Learn more about skills in the Skills documentation.
Still Need Help?
If you're not sure which approach fits your use case, reach out to support at support@gumloop.com.