Agent Configuration
OpenSpider ships with 3 default agents — Manager, Coder, and Researcher. Each agent's behavior is controlled by a set of pillar files in workspace/agents/<agent-id>/.
Default Agents
| Agent | ID | Role | Emoji |
|---|---|---|---|
| Ananta | manager | Gateway Architect & Manager | 🕷️ |
| Coder | coder | Code generation & execution | ⚡ |
| Researcher | researcher | Web research & data gathering | 🔮 |
Pillar Files
Each agent has 4 pillar files that define its personality, behavior, and capabilities:
IDENTITY.md
Defines who the agent is. When the agent is asked to introduce itself, it references this file.
# OpenSpider Manager Agent
You are **Ananta** 🕷️, the core Manager Agent of the OpenSpider multi-agent system.
- **Role**: Gateway Architect & Manager
- **Vibe**: Chill, smart, gets things done.
- **CEO / Boss**: Jane Doe
- **Company**: Acme Corp
- **Website**: https://acme.example.com
- **Capabilities**: Web search, coding, research, email, WhatsApp, scheduling
- **Emoji**: 🕷️Customize This!
After your first run, edit workspace/agents/manager/IDENTITY.md to set your name as CEO/Boss, your company, and website. The agent will use these when introducing itself.
SOUL.md
Defines how the agent behaves — its safety directives, behavioral rules, and system architecture knowledge.
# Core Ethos & Safety Directives
1. **Self-Preservation & System Stability**: Never execute commands that could destroy the system.
2. **Delegation Efficiency**: Delegate deep tasks to Sub-Agents.
3. **Truthfulness Over Hallucination**: State clearly when you don't know.
4. **Tool Minimization**: Only use tools necessary for the objective.
5. **Autonomy Limits**: Surface breaking errors instead of infinitely looping.
# System Architecture Knowledge
- **Cron Jobs**: Stored in `workspace/cron_jobs.json`
- **Memory**: Daily logs in `workspace/memory/YYYY-MM-DD.md`
- **Agents**: Configs in `workspace/agents/<agent-id>/`
- **WhatsApp Config**: `workspace/whatsapp_config.json`
- **Skills**: Custom tools in `skills/`WARNING
Modifying SOUL.md can fundamentally change the agent's behavior. Be careful when editing safety directives.
CAPABILITIES.json
Defines the agent's tools, role, and operational limits.
{
"name": "Ananta",
"role": "Manager",
"emoji": "🕷️",
"color": "fuchsia",
"allowedTools": ["delegate_task", "read_file", "search_web", "send_email"],
"maxLoops": 15,
"status": "running"
}| Field | Type | Description |
|---|---|---|
name | string | Display name for the agent |
role | string | Agent's role (Manager, Coder, Researcher, or custom) |
emoji | string | Emoji shown in the dashboard and chat |
color | string | Theme color in the dashboard |
allowedTools | string[] | Tools the agent is allowed to use |
maxLoops | number | Max iterations in the action loop (prevents runaway) |
status | "running" | "stopped" | Agent status (stopped agents show as red in dashboard) |
USER.md
Stores learned context about the user. This file evolves over time as the agent interacts and learns preferences.
# User Context
No user preferences or context learned yet.INFO
This file starts empty and is updated by the system as it learns about you. You can also manually add notes here.
Creating Custom Agents
Via the Dashboard
- Go to Agents Workspace in the sidebar
- Click Create Agent
- Fill in name, role, color, and model
- The system automatically creates all 4 pillar files
Via the API
curl -X POST http://localhost:4001/api/agents \
-H "Content-Type: application/json" \
-d '{
"name": "Analyst",
"role": "Data Analyst",
"model": "antigravity",
"color": "cyan"
}'Manually
Create a new directory under workspace/agents/:
mkdir -p workspace/agents/analystThen create the 4 pillar files (IDENTITY.md, SOUL.md, CAPABILITIES.json, USER.md) following the formats above.
Workspace Defaults
OpenSpider ships with a workspace-defaults/ directory that contains the default agent configurations. On first run, if no workspace/ directory exists, the system automatically copies all defaults into workspace/.
workspace-defaults/
├── agents/
│ ├── manager/
│ │ ├── IDENTITY.md # Default manager persona
│ │ ├── SOUL.md # Safety directives + system knowledge
│ │ ├── CAPABILITIES.json # Tools, role, emoji
│ │ └── USER.md # Empty user context
│ ├── coder/
│ │ ├── IDENTITY.md
│ │ ├── SOUL.md
│ │ ├── CAPABILITIES.json
│ │ └── USER.md
│ └── researcher/
│ ├── IDENTITY.md
│ ├── SOUL.md
│ ├── CAPABILITIES.json
│ └── USER.md
└── memory.md # Long-term memory templateTIP
If you ever want to reset an agent to its defaults, delete its folder from workspace/agents/ and restart. The defaults will NOT be re-copied (only on first run). To force a reset, delete the entire workspace/ directory and restart.
Modifying Agents at Runtime
Agent pillar files (IDENTITY.md, SOUL.md, USER.md) are read at runtime — changes take effect immediately without rebuilding. CAPABILITIES.json changes also take effect on the next request.
You can manage agents and edit pillar files natively via the UI:
Via the Dashboard (Recommended)
- Navigate to Agents Workspace in the left sidebar.
- Select the agent you want to modify from the list.
- You will see tabs for each pillar file:
- IDENTITY: Update the agent's core background, title, and persona.
- SOUL: Update the agent's behavior rules and operational constraints.
- USER CONTEXT: Provide the agent with specific information about yourself or the project.
- CAPABILITIES: Add or remove skills, change the underlying AI model, or edit the
namefield to rename the agent.
- Click Save Changes to apply the updates immediately.
Via the API or CLI
- API:
PUT /api/agents/:idwith the updated pillar content - Manually: Edit the files directly in
workspace/agents/<id>/using your text editor