Agents
An agent in Snippbot is a persistent AI identity with its own personality, memory, tools, and workspace. You can have multiple agents, each configured differently for different purposes.
Agent properties
Section titled “Agent properties”| Property | Description |
|---|---|
id | Unique identifier (agt_ prefix) |
name | Display name (e.g. “Aria”, “Atlas”) |
archetype | Personality template |
model | AI provider + model |
system_prompt | Custom system instructions |
memory_enabled | Whether to use episodic memory |
knowledge_graph | Whether to extract entities from memory |
workspace_dir | Private file workspace path |
trust_level | Permission tier (Observer → Admin) |
voice | TTS voice configuration |
avatar | Avatar image |
Archetypes
Section titled “Archetypes”Archetypes are personality templates that configure default traits, communication style, and tool preferences:
| Archetype | Description | Best for |
|---|---|---|
| Personal Companion | Friendly, empathetic, conversational | Daily assistance, casual interaction |
| Professional Assistant | Organized, efficient, task-focused | Business tasks, scheduling, email |
| Researcher | Methodical, thorough, cites sources | Research, analysis, fact-checking |
| Teacher / Mentor | Patient, explanatory, encouraging | Learning, tutoring, skill-building |
| Creative Partner | Imaginative, collaborative, aesthetic | Writing, design, brainstorming |
| Coach / Motivator | Supportive, goal-oriented, accountable | Fitness, habits, personal growth |
| Technical Expert | Precise, systems-focused, detail-oriented | Code, architecture, debugging |
| Entrepreneur Partner | Strategic, action-oriented, decisive | Business planning, startups |
| Health / Wellness | Caring, holistic, science-informed | Health tracking, wellness routines |
| Accountability Partner | Direct, consistent, progress-tracking | Goals, deadlines, commitments |
| Thinking Partner | Analytical, Socratic, nuanced | Problem-solving, decision-making |
| Something Unique | Fully customizable | Any custom use case |
An archetype affects the agent’s system prompt, but you can fully customize the system prompt afterwards.
Model configuration
Section titled “Model configuration”Each agent is assigned a provider and model:
Snippbot supports 10 providers:
| Provider | Auth type | Notable models |
|---|---|---|
| Claude (Native Tools) | Subscription (Claude CLI) | Claude Opus 4.6, Sonnet 4.6, Sonnet 4.5, Haiku 4.5 |
| Anthropic API | API key | Claude Opus 4.6, Sonnet 4.6, Sonnet 4.5, Haiku 4.5 |
| OpenAI | API key | GPT-5.3 Codex, GPT-5.2, GPT-4o, o3, o4-mini |
| OpenRouter | API key | 400+ models (Claude, GPT, Gemini, Llama, DeepSeek, Grok) |
| Google Gemini | API key | Gemini 2.5 Pro, 2.5 Flash, 3.1 Pro Preview, 3 Flash Preview |
| DeepSeek | API key | DeepSeek V3.2, DeepSeek R1 (Reasoner) |
| xAI Grok | API key | Grok 4.20, Grok 4, Grok 4 Fast, Grok 3 |
| Groq | API key | GPT-OSS 120B, GPT-OSS 20B, Llama 4 Scout, Llama 3.3 70B |
| Mistral AI | API key | Mistral Large 3, Mistral Small 4, Devstral 2, Codestral |
Configure the provider and model from the Agents detail page in the UI under the Model section.
All 10 supported providers share a unified interface — switching an agent between providers doesn’t require changing task definitions.
Memory
Section titled “Memory”Each agent has access to a memory system with multiple layers:
| Layer | Description | Search method |
|---|---|---|
| Episodic memory | Time-stamped event log | Keyword (FTS5) |
| Vector store | Embedding-indexed memories | Semantic similarity |
| Knowledge graph | Entities and relationships | Graph traversal |
Memory is scoped per agent — agents don’t share memory by default. See Memory concepts for the full memory model.
Workspace
Section titled “Workspace”Each agent has a private workspace directory (~/.snippbot/workspaces/{agent_id}/) for file storage. Tasks write output files here, and the agent can read and modify them across sessions.
The workspace is accessible via:
- The Agents → Workspace tab in the UI
- Agent tools (
write_file,read_file,list_files)
Sub-agents
Section titled “Sub-agents”Agents can spawn sub-agents for parallel or specialized work. Sub-agents have 7 roles:
| Role | Description |
|---|---|
researcher | Information gathering and research |
coder | Code writing and script execution |
reviewer | Quality assurance and code review |
tester | Testing and validation |
analyst | Data analysis and insights |
creative | Content creation and design |
general | General-purpose tasks |
Sub-agent nesting is limited to 3 levels to prevent infinite recursion. See Use Sub-Agents.
Related
Section titled “Related”-
First Agent — create and configure your first agent
-
Memory — how agents remember things
-
Use Sub-Agents — multi-agent orchestration
