Agent Configuration

This reference covers every configuration option available for agents on systemprompt.io. Use this page when you need to understand what a setting does or how to fine-tune agent behavior.

Basic settings

These settings define the agent's identity and are required for every agent.

Setting	Required	Description
Name	Yes	A short, descriptive label for the agent. Displayed in the dashboard and used by the platform to identify the agent. Keep it under 50 characters.
Description	Yes	A one or two sentence summary of the agent's purpose. The platform uses this to help route requests to the correct agent. Be specific about what the agent does.
Status	Yes	Whether the agent is Active or Inactive. Inactive agents are not available through the MCP server and will not handle any requests.

Tips for name and description

Your agent's description directly affects how well the platform matches incoming requests. Compare these two descriptions:

• Weak: "Helps with writing tasks."
• Strong: "Writes and edits blog posts, landing page copy, and email newsletters for our B2B SaaS product."

The strong description gives the platform clear signals about what this agent handles, which reduces misrouted requests.

System prompt

The system prompt is the core of your agent's behavior. It tells Claude how to act when operating as this agent.

Setting	Required	Description
System prompt	Yes	Instructions that define the agent's persona, tone, boundaries, and operational rules. This is sent to Claude as context for every interaction with this agent.

Writing effective system prompts

A good system prompt covers four areas:

1. Role: What the agent is and what it does. ("You are a technical documentation writer.")
2. Tone: How the agent communicates. ("Write in a clear, direct style. Avoid marketing language.")
3. Boundaries: What the agent should not do. ("Do not provide legal advice or make commitments on behalf of the company.")
4. Context: Background knowledge the agent should always have. ("Our product is a cloud-based project management tool for teams of 10-50 people.")

Keep system prompts between 5 and 20 lines. Shorter prompts are easier to maintain and produce more consistent behavior. If you need more than 20 lines, consider whether some of that content belongs in a skill instead.

Example system prompt

You are a technical writer for systemprompt.io. Write clear, concise documentation
that helps developers understand and use the platform.

Use short sentences and active voice. Include code examples when relevant.
Structure content with headings, numbered steps, and bullet points.

Do not assume the reader's experience level. Define technical terms on first use.
Do not reference internal tools or processes that customers cannot access.

Skills assignment

Skills define the specific tasks your agent can perform. An agent with no skills has no structured capabilities.

Setting	Required	Description
Assigned skills	No (but recommended)	The list of skills this agent can use. Each skill provides a defined task with its own instructions and parameters.

How skill assignment works

When you assign a skill to an agent, that skill becomes available as a tool through the MCP server. Claude sees the skill's name, description, and parameters, and uses them to handle matching requests.

You can assign any number of skills to a single agent. However, keep these guidelines in mind:

• Group related skills. An agent with "Write Blog Post," "Write Newsletter," and "Edit Copy" makes sense. An agent with "Write Blog Post" and "Run Database Query" does not.
• Avoid overlap. If two skills have very similar descriptions, Claude may struggle to pick the right one. Make skill descriptions distinct.
• Start small. Begin with 2-3 skills and add more as you identify gaps in the agent's capabilities.

Managing skills

To add or remove skills from an agent:

1. Open the agent in the dashboard.
2. Go to the Skills section.
3. Click Add Skill to browse and select available skills.
4. To remove a skill, click the remove icon next to the skill name.
5. Click Save to apply changes.

Changes to skill assignments take effect immediately. You do not need to restart the MCP server.

Plugin assignment

Plugins extend your agent with access to external tools and services through the MCP protocol.

Setting	Required	Description
Assigned plugins	No	The list of plugins this agent can use. Plugins provide MCP tools for interacting with external systems, file storage, APIs, and other services.

When to add plugins

Add plugins when your agent needs to do more than generate text. Common scenarios:

• File access: The agent needs to read or write files.
• Data retrieval: The agent needs to query a database or API.
• External actions: The agent needs to create records, send notifications, or trigger workflows in other systems.

If your agent only needs to produce text output (writing, analysis, answering questions), it may not need any plugins at all.

Plugin permissions

Each plugin has its own set of capabilities. When you assign a plugin to an agent, the agent gains access to all of that plugin's tools. Review the plugin's documentation to understand exactly what access it provides.

Advanced settings

These settings give you finer control over agent behavior.

Setting	Required	Default	Description
Context window	No	Platform default	Controls how much conversation history is included when the agent processes a request. Larger context windows let the agent reference earlier parts of a conversation but consume more tokens.
Temperature	No	Platform default	Controls the randomness of the agent's output. Lower values (closer to 0) produce more predictable, focused responses. Higher values produce more varied and creative responses.
Max output length	No	Platform default	The maximum number of tokens the agent can generate in a single response. Set this to prevent unexpectedly long responses.
Tags	No	None	Labels for organizing and filtering agents in the dashboard. Tags do not affect agent behavior.

Context window guidance

For most agents, the platform default context window works well. Consider adjusting it in these cases:

• Increase for agents that handle long, multi-turn conversations where referencing earlier context is important (support agents, research agents).
• Decrease for agents that handle short, independent tasks where earlier context is irrelevant (one-off content generation, simple Q&A).

Temperature guidance

• 0.0 - 0.3: Best for factual, consistent output. Use for support agents, documentation writers, data analysis.
• 0.4 - 0.7: Balanced creativity and consistency. Use for general content writing, brainstorming with some structure.
• 0.8 - 1.0: More creative and varied. Use for creative writing, ideation, or tasks where diversity of output is valued.

Configuration checklist

Use this checklist when setting up or reviewing an agent:

1. Name is specific and under 50 characters.
2. Description clearly states what the agent does (not what it is).
3. System prompt covers role, tone, boundaries, and context.
4. Skills are relevant, complementary, and have distinct descriptions.
5. Plugins are added only where external access is needed.
6. Advanced settings are adjusted only if the defaults do not meet your needs.
7. Agent status is set to Active when ready for use.

Troubleshooting

Agent produces inconsistent responses. Check the temperature setting. Lower it for more predictable output. Also review the system prompt for ambiguous instructions that could be interpreted multiple ways.

Agent ignores system prompt instructions. Make sure the system prompt is concise and clear. Conflicting instructions cause unpredictable behavior. If the prompt is longer than 20 lines, simplify it and move detailed task instructions into skills instead.

Agent cannot access expected tools. Verify that the correct plugins are assigned. Check that the plugins are active and properly configured. Review the plugin's logs for errors.

Changes to agent settings are not reflected. Confirm that you clicked Save after making changes. If using Claude Desktop, try disconnecting and reconnecting the MCP server to pick up configuration changes.

Agent handles requests meant for a different agent. Improve the description for both agents to make their roles more distinct. Vague or overlapping descriptions cause routing confusion.

Related pages

• Creating Agents -- Step-by-step guide to building an agent.
• Skills -- Create the skills your agents will use.
• Plugins -- Connect agents to external tools and services.