Create an AI Chat Agent
An AI Chat Agent is a specialized agent with its own instructions, knowledge, and capabilities. It gives you a reusable way to handle one focused task from request to review-ready answer.
Use a Chat Agent when you want Marc to apply a consistent role, process, or set of capabilities in a conversation. For example, you can create Chat Agents for financial analysis, translation, document review, market research, or internal knowledge support.
A specialized Chat Agent saves the setup you need for a recurring type of task. Prepare it once, then select it in chat whenever you need that expertise.
What can a Chat Agent include?
A Chat Agent can save:
- Instructions - the role, tone, boundaries, and response format Lampi should follow
- A default user message - a message automatically added when you select the Chat Agent with
@ - Files and folders - knowledge sources the Chat Agent should use by default
- Tools - capabilities such as Files Search, Web Search, or workspace-specific tools
- Skills - reusable workflows that guide Lampi through a specific task
- Research Mode - whether the Chat Agent should use the chat default, Tooling, or Deep Search
- Model settings - the model and generation behavior used by the Chat Agent
Create a Chat Agent
To create a custom Chat Agent, open Settings > AI Chat Agents, select Create Chat Agent, and follow these steps.
1. Choose the Chat Agent type and name
Choose who can use the Chat Agent:
- Personal - available only to your account
- Team - available to members of your organization, when enabled by an admin
- Public - available more broadly, when enabled by an administrator
Then give the Chat Agent a clear name, such as "Financial Analyst" or "Contract Review Specialist."
Lampi automatically creates an @ trigger from the Chat Agent name. For example, a Chat Agent named "Financial Analyst" can be selected in chat with @financial-analyst.
You can also add a short description so other users understand when to use the Chat Agent.
2. Add a Default User Message
The User Message is the message Lampi automatically adds to the chat when you select the Chat Agent with @.
Use this when the Chat Agent should always start from the same request or template. For example:
Review the attached document and return:
1. Key takeaways
2. Risks or inconsistencies
3. Follow-up questions
You can also select Load from databank to attach files or images to this default message. These attachments are added to the chat together with the message when the Chat Agent is selected.
Use the User Message for the task you want completed. Use the Chat Agent instructions for how the work should be performed.
3. Write Instructions
Instructions define how the Chat Agent works. This is where you describe its role, style, boundaries, and working method.
When you use a Chat Agent, Lampi can use two kinds of text:
- Instructions - long-term guidance for how the Chat Agent should work
- User Message - the message or task added to chat when the Chat Agent is selected
Guidelines for Writing Instructions
Give your Chat Agent a role.
A clear role helps Lampi answer from the right perspective. For example, "You are an experienced financial analyst" is more useful than "Help me with finance."
Provide detailed guidance, including:
- Tone of communication - for example, professional, concise, analytical, or client-ready
- Source usage - explain whether the Chat Agent should rely only on selected files, use web information, or combine both when available
- Company-specific context - include internal rules, naming conventions, or preferred terminology
- Response format - specify whether answers should use bullets, tables, executive summaries, citations, or a fixed template
- Special cases - tell Lampi what to do when information is missing, uncertain, or outside the selected sources
- Working method - describe the steps the Chat Agent should follow for recurring tasks
If several Chat Agents need the same specialized workflow, create a reusable skill instead of copying long instructions into each one. For more, see Skills.
Examples of Instructions
Content Writer
You are an expert content writer skilled in creating engaging, original, and impactful content.
You can assist in brainstorming ideas, developing outlines, crafting compelling introductions and conclusions, and refining content for clarity and coherence. You are knowledgeable about SEO, audience engagement strategies, and different writing styles to match the specific needs of various types of content. You have extensive experience in writing SEO-optimized blog posts.
Use an engaging and informative tone of voice. Format your response using markdown and structure your responses with clear tips and suggestions to make them easy to implement.
If you are uncertain about your answer, clearly state that the information is not available and ask the user to clarify or provide additional details.
📊 Financial Analyst
You are an experienced financial analyst, specializing in providing detailed, insightful, and data-driven information.
You can assist with financial analysis, investment evaluation, and risk management. Your expertise extends to understanding economic trends, financial ratios, and data interpretation to provide comprehensive insights and strategies.
Ground your conclusions in the information and sources available to the assignment. Cite supporting evidence, distinguish facts from interpretation, and clearly identify missing or uncertain information.
Adopt a formal and analytical tone. Format your response using markdown and organize your responses with structured paragraphs to ensure clarity.
If you are uncertain about your answer, clearly state that the information is not available and ask the user to clarify or provide additional details.
Grammar Specialist
You are a grammar specialist dedicated to enhancing the quality and precision of written communication.
Your task is to take the text provided and rewrite it into a clear, grammatically correct version while preserving the original meaning as closely as possible. Correct grammatical and punctuation errors, spelling mistakes, verb tense issues, suggesting stylistic improvements, and ensuring clarity and conciseness. You are well-versed in grammar rules, punctuation, and syntax.
Each answer should contribute to producing polished, error-free text.
Always start with the corrected sentence and continue with a very concise note on the changes made and why.
Translation Specialist
You are a translation specialist focused on providing accurate, context-aware translations between languages.
You can support with translating documents, maintaining cultural nuances, and suggesting language adjustments for clarity and tone. You have a deep understanding of idiomatic expressions, formal and informal registers, and technical terminology across multiple languages.
Your answers should be accurate and maintain the intended meaning of the original text. Each response should be a trusted source for high-quality translation support.
4. Add Knowledge Sources
Use Knowledge & Tools to decide what the Chat Agent can rely on by default.
Select Specific Files
Choose individual files when the Chat Agent should always use a defined set of documents. This is useful for agents dedicated to a policy, report, contract, knowledge base, or recurring client file.
For example, a "Contract Review" Chat Agent might include your standard contract playbook and approved clause library.
Synchronize Folders
Choose folders when the Chat Agent should have access to a broader collection of documents. This is useful when the relevant knowledge changes over time or when the agent needs to search across a team folder.
Keep knowledge sources focused. A Chat Agent with a few relevant files or folders usually performs better than one with too much unrelated context.
5. Select Tools and Skills
Tools and skills give the Chat Agent capabilities beyond its instructions.
Tools
Tools let the Chat Agent perform actions or access information sources. Depending on your workspace, this can include Files Search, Web Search, connected services, specialized databases, or custom tools.
Use tools when the Chat Agent needs to retrieve information, work with a connection, or perform an available action.
Skills
Skills are reusable workflows. They tell Lampi how to approach a specific type of work, such as researching a company, reviewing a document, creating slides, or producing a structured report.
Add skills to a Chat Agent when you want that workflow available by default every time the agent is selected.
For more, see Skills and Tools & MCP Library.
6. Choose Research Mode
Research Mode controls how Lampi should run the Chat Agent when it is selected in chat.
| Mode | What It Does | Best For |
|---|---|---|
| Default | Uses the mode selected in chat for that request. | Flexible Chat Agents that should adapt to the current chat setup. |
| Tooling | Always runs the Chat Agent with its selected tools and skills. | Work that should consistently use saved capabilities, such as file search, web search, or workspace tools. |
| Deep Search | Always uses the deeper research workflow when the Chat Agent is selected. | Complex research, comparisons, summaries, diligence work, and evidence-backed reports. |
Deep Search is designed for deeper synthesis than a standard answer. It can plan research, gather information from selected sources, compare findings, and produce a more complete result.
To select Deep Search, first add a search capability such as Files Search or Web Search to the Chat Agent.
7. Select the Model
Choose the model that will power the Chat Agent.
Lampi AI is designed to provide a secure and confidential way to interact with AI and your knowledge. When you select a model hosted by your company or Lampi, the data you provide stays secure and under control. If you select a model hosted by another provider, some data may be sent to that provider to process your request.
The best model depends on the task. Test the Chat Agent on realistic examples before using it for important work.
8. Adjust Advanced Model Settings
Advanced settings control how the model generates answers in different situations.
Temperature
Temperature controls how predictable or creative the answer is.
| Setting | Characteristics | Use Cases |
|---|---|---|
| Low Temperature | More predictable, consistent, and conservative | Factual answers, document review, translations, compliance-sensitive work |
| High Temperature | More varied, exploratory, and creative | Brainstorming, drafting alternatives, creative writing |
Temperature (Tooling)
Temperature (Tooling) controls generation behavior when the Chat Agent is using tools or skills. Lower values are usually better for precise tool use, accurate extraction, or reliable source-based answers.
Temperature (Summary)
Temperature (Summary) controls generation behavior for summary-style steps. Lower values are usually better for concise factual summaries. Higher values can be useful when you want more varied wording or synthesis.
There is no universal best temperature. Use lower values for reliability and higher values for creative or exploratory tasks.
Max Generation Tokens
Max generation tokens controls the maximum length of the Chat Agent's answer. If the answer reaches the limit, it may stop before completing.
You cannot increase the limit beyond what the selected model supports. For longer deliverables such as reports, consider using Deep Search or AI Grid Agents.
Use a Chat Agent
In chat, type @ and select the Chat Agent you want to use. Lampi adds it to the message and applies its saved configuration.
When a Chat Agent is selected, Lampi can use:
- The Chat Agent's instructions
- Its default user message and databank attachments
- Its selected files and folders
- Its tools and skills
- Its Research Mode
- Its model settings
You can still write additional instructions in the chat message. The Chat Agent provides the default setup, and your message defines what you need now.
Next: Troubleshooting →