CLI Reference - PraisonAI

First-run Credential Check

Both praisonai (bare) and praisonai run verify credentials before executing agent work. Detected env vars (any one satisfies the check): any credential key in PROVIDER_ENV_CATALOGUE — OPENAI_API_KEY, ANTHROPIC_API_KEY, GOOGLE_API_KEY, GEMINI_API_KEY, GROQ_API_KEY, COHERE_API_KEY, OPENROUTER_API_KEY, OLLAMA_HOST, MISTRAL_API_KEY, DEEPSEEK_API_KEY, XAI_API_KEY, TOGETHER_API_KEY / TOGETHERAI_API_KEY, PERPLEXITYAI_API_KEY, FIREWORKS_API_KEY / FIREWORKS_AI_API_KEY (full 14-row table). Stored credentials from praisonai setup are also accepted. See First-run Onboarding for details.

Command Tree

praisonai
├── [direct prompt]              # Any text → runs agent (in-process on standalone as of PR #2853)
├── [file.yaml]                  # YAML workflow execution
├── praisonai chat           # TUI mode (resident standalone; wrapper adds richer dispatch)
├── praisonai chat                  # Single prompt chat mode (resident standalone; wrapper adds richer dispatch)
│
├── chat                         # Chainlit chat UI (port 8084)
├── code                         # Chainlit code UI (port 8086)
├── call                         # PraisonAI Call server
│   └── --port --host --public
├── realtime                     # Realtime voice UI (port 8088)
├── train                        # Model training
├── ui                           # Gradio/Chainlit UI (port 8082)
│
├── context                      # Context engineering
│   └── --url, --goal, --auto-analyze
├── research                     # Deep research agent
│   └── --query-rewrite, --tools (file path OR comma-separated names via full resolver), --save
│
├── checkpoint                   # File-level shadow-git checkpoints
│   ├── save <msg>               # Snapshot the workspace
│   │   └── --allow-empty, -w
│   ├── list                     # Show recent checkpoints (newest first)
│   │   └── --limit/-n, -w
│   ├── restore <id|last>        # Rewind workspace to a checkpoint
│   │   └── -w
│   ├── diff [from] [to]         # Diff between checkpoints or working dir
│   │   └── -w
│   └── delete                   # Delete all checkpoints
│       └── --yes/-y, -w
│
├── memory                       # Memory management
│   ├── show                     # Show current memory (--limit N, --user-id ID)
│   ├── list                     # Alias for show (--limit N, --user-id ID)
│   ├── add <content>            # Add memory entry
│   ├── search <query>           # Search memories
│   ├── clear                    # Clear all memories
│   ├── save <name>              # Save session
│   ├── resume <name>            # Resume session
│   ├── sessions                 # List sessions
│   ├── compress                 # Compress memory
│   ├── checkpoint               # Create memory checkpoint
│   ├── restore <id>             # Restore memory checkpoint
│   └── checkpoints              # List memory checkpoints
│
├── rules                        # Rules management
│   ├── list                     # List all rules
│   ├── show <name>              # Show specific rule
│   ├── create <name> <content>  # Create rule
│   ├── delete <name>            # Delete rule
│   └── stats                    # Rule statistics
│
├── workflow                     # Workflow management
│   ├── list                     # List workflows
│   ├── run <file>               # Run workflow
│   ├── show <file>              # Show workflow details
│   ├── create                   # Create workflow
│   ├── validate <file>          # Validate workflow
│   ├── template <name>          # Create from template
│   └── auto <topic>             # Auto-generate workflow
│
├── flow                         # Visual workflow builder (Langflow)
│   ├── (no args)                # Start the builder UI
│   ├── import <file>            # Import YAML → Langflow
│   ├── export <flow_id>         # Export Langflow → YAML/JSON
│   ├── list                     # List flows on a server
│   └── version                  # Show Langflow version
│
├── hooks                        # Hooks management
│   ├── list                     # List hooks
│   ├── stats                    # Hook statistics
│   └── init                     # Create hooks.json
│
├── knowledge                    # Knowledge/RAG management
│   ├── add <source>             # Add knowledge source
│   ├── query <query>            # Query knowledge
│   ├── list                     # List sources
│   ├── clear                    # Clear knowledge
│   └── stats                    # Knowledge statistics
│
├── session                      # Session management
│   ├── start                    # Start new session
│   ├── list                     # List sessions
│   ├── resume <id> [PROMPT]     # Resume session with full state restore
│   │   └── --transcript         # Show transcript only (no state restore)
│   ├── delete <id>              # Delete session
│   └── info <id>                # Session info
│
├── usage                        # Local token/cost usage reporting
│   └── --days/-d --by/-b --project/-p --json
│
├── tools                        # Tool management
│   ├── list                     # List available tools
│   ├── info <name>              # Tool information
│   └── search <query>           # Search tools
│
├── todo                         # Todo management
│   ├── list                     # List todos
│   ├── add <content>            # Add todo
│   ├── complete <id>            # Complete todo
│   ├── delete <id>              # Delete todo
│   └── clear                    # Clear all todos
│
├── docs                         # Documentation management
│   ├── run                      # Run doc code validation
│   ├── list                     # List docs/code blocks
│   ├── stats                    # Show group statistics
│   ├── run-all                  # Run all groups
│   ├── report [path]            # View execution report
│   │   └── --limit, --wide, --match, --group, --format
│   ├── cli                      # CLI command validation
│   │   ├── run-all              # Validate all CLI commands
│   │   ├── list                 # List CLI commands
│   │   ├── stats                # CLI command statistics
│   │   └── report               # View CLI validation report
│   ├── api-md                   # Generate API reference (api.md)
│   │   └── --write, --check, --stdout
│   ├── generate                 # Generate documentation
│   └── serve                    # Serve docs locally
│
├── examples                     # Examples management
│   ├── run                      # Run examples
│   ├── list                     # List examples
│   ├── stats                    # Show group statistics
│   ├── run-all                  # Run all groups
│   └── report [path]            # View execution report
│       └── --limit, --wide, --match, --group, --format
│
├── mcp                          # MCP server management
│   ├── list                     # List MCP configs
│   ├── show <name>              # Show config
│   ├── create <name> <cmd>      # Create config
│   ├── delete <name>            # Delete config
│   ├── enable <name>            # Enable config
│   └── disable <name>           # Disable config
│
├── commit                       # AI commit message generation
│   └── --push, -a/--auto, --no-verify
│
├── serve                        # API server
│   └── <agents.yaml> --port --host
│
├── schedule                     # Task scheduling
│   ├── start                    # Start scheduler
│   ├── list                     # List jobs
│   ├── stop <id>                # Stop job
│   ├── logs <id>                # View logs
│   ├── restart <id>             # Restart job
│   ├── delete <id>              # Delete job
│   ├── describe <id>            # Job details
│   ├── save                     # Save state
│   ├── stop-all                 # Stop all jobs
│   └── stats                    # Scheduler stats
│
├── skills                       # Agent Skills management
│   ├── list                     # List skills
│   ├── validate <path>          # Validate skill
│   ├── create <name>            # Create skill
│   └── install <repo>           # Install skill
│
├── profile                      # Profiling
│   └── <prompt>                 # Profile agent execution
│
├── eval                         # Evaluation framework
│   ├── accuracy                 # Accuracy evaluation
│   ├── performance              # Performance benchmark
│   ├── reliability              # Tool reliability check
│   └── criteria                 # Custom criteria eval
│
├── doctor                       # Health checks & diagnostics
│   ├── env                      # Environment checks
│   ├── config                   # Configuration validation
│   ├── tools                    # Tool availability
│   ├── db                       # Database checks
│   ├── mcp                      # MCP configuration
│   ├── obs                      # Observability providers
│   ├── skills                   # Agent skills
│   ├── memory                   # Memory storage
│   ├── permissions              # Filesystem permissions
│   ├── network                  # Network connectivity
│   ├── performance              # Import times
│   ├── ci                       # CI mode
│   └── selftest                 # Agent functionality
│
├── agent                        # Custom agent definitions
│   ├── create [name]            # Author a new custom agent (interactive or --yes)
│   ├── list   [--verbose]       # List discovered custom agents
│   └── show   <name>            # Show agent details
│
├── auth                         # Credential management
│   ├── login <provider>         # Store provider key (interactive or --key)
│   ├── list                     # List stored providers (redacted)
│   ├── status [provider]        # Validate stored credentials
│   └── logout [provider|--all]  # Remove credentials
│
├── command                      # Custom command templates
│   ├── list [--verbose]         # List discovered commands
│   └── show <name> [--preview]  # Show / preview a command template
│
├── models                       # LLM model catalogue
│   ├── list [SEARCH]            # List/filter models by provider or name
│   ├── describe <model>         # Full capabilities/limits/cost
│   └── validate <model>         # Validate an ID; suggests alternatives on miss
│
├── permissions                  # Tool approval rules
│   ├── list                     # List rules
│   ├── allow|deny|ask <pattern> # Add a rule
│   ├── remove <id-prefix>       # Remove a rule
│   ├── reset                    # Delete all rules (confirm)
│   ├── export                   # Print rules as JSON
│   └── import <file>            # Import rules from JSON
│
├── validate                     # YAML configuration validation
│   ├── <file>                   # Validate one file
│   ├── check [directory]        # Validate all YAML in directory
│   └── schema                   # Print the YAML schema
│
├── agents                       # Agent management
├── run                          # Run agents
├── thinking                     # Thinking budget config
├── compaction                   # Context compaction config
├── output                       # Output style config
│
├── deploy                       # Deployment management
│   ├── init                     # Initialize deployment
│   ├── validate                 # Validate config
│   ├── plan                     # Show deployment plan
│   ├── status                   # Deployment status
│   ├── destroy                  # Destroy deployment
│   ├── run                      # Run deployment
│   ├── api                      # API deployment
│   ├── docker                   # Docker deployment
│   └── cloud                    # Cloud deployment
│
├── templates                    # Template management
│
└── [Capabilities - LiteLLM parity] (27 commands)
    ├── audio                    # Audio transcription/TTS
    ├── embed                    # Embeddings
    ├── images                   # Image generation
    ├── moderate                 # Content moderation
    ├── files                    # File management
    ├── batches                  # Batch processing
    ├── vector-stores            # Vector store management
    ├── rerank                   # Reranking
    ├── ocr                      # OCR
    ├── assistants               # Assistants API
    ├── fine-tuning              # Fine-tuning
    ├── completions              # Completions
    ├── messages                 # Messages
    ├── guardrails               # Guardrails
    ├── rag                      # RAG
    ├── videos                   # Video processing
    ├── a2a                      # Agent-to-Agent
    ├── containers               # Container management
    ├── passthrough              # Passthrough requests
    ├── responses                # Response management
    ├── search                   # Search
    └── realtime-api             # Realtime API

praisonai agent create authors a new .praisonai/agents/<name>.md definition interactively or with --yes. See Create Custom Agents for the full walkthrough.

Direct Prompt Examples

The [direct prompt] entry in the tree above means a bare positional that isn’t a .yaml/.yml name routes to Typer run — the same modern engine as praisonai run "<prompt>". Flags accepted by run itself (--model, --continue, --session, --output, --stream, …) are forwarded to run intact; only a genuinely legacy-only flag (e.g. --framework, --auto, --serve) keeps the invocation on the legacy dispatcher (and now prints a one-line stderr notice when it does):

praisonai "summarise this folder"   # bare prompt → Typer run (modern engine)
praisonai show                      # bare token → Typer run, prompt "show"
praisonai memoyr                    # close typo of memory → exit 2, Did you mean: memory?
praisonai agents.yaml               # flagless .yaml → Typer run (modern engine)
praisonai memory show               # known subcommand → routed normally

praisonai show --framework crewai   # flag → legacy; ⚠ reserved-verb hint, exit 2

Two typo guards apply. The mistyped-verb guard on the modern dispatcher blocks a lone, non-prefix close typo of a real command (praisonai deploi, memoyr) with exit 2 and a Did you mean …? suggestion, on any invocation. The reserved-verb guard on the legacy path fires only once a legacy-only flag has already forced legacy. A plain praisonai show is not a close match, so it routes to Typer run and is sent to the model as the literal prompt show. See CLI Dispatcher → Two typo guards.

A flagless .yaml/.yml positional reaches the modern run engine — praisonai agents.yaml is equivalent to praisonai run agents.yaml, inheriting session continuity, --output modes, the credential gate and permissions. Global flags in front of it are hoisted ahead of run (praisonai --output-format json agents.yaml → --output-format json run agents.yaml). Only a genuinely legacy-only flag (--auto, --serve, --n8n, -s, -f) keeps it on the legacy roles-file dispatcher. See CLI Dispatcher.

Global Flags (70+ flags)

These root-callback flags belong before the subcommand — praisonai --output-format json version, not praisonai version --output-format json. Click accepts a group-level option only ahead of the subcommand; in front of a bare prompt or .yaml target the dispatcher hoists them ahead of the synthesised run. -o / -v are shared with run and stay with it.

Flag	Short	Type	Description
`--output-format`	`-o`	choice	`text` \| `json` \| … output format; env `PRAISONAI_OUTPUT_FORMAT`
`--json`	—	flag	Alias for `--output-format json`
`--no-color`	—	flag	Disable colored output; env `NO_COLOR`
`--quiet`	`-q`	flag	Minimal output
`--screen-reader`	—	flag	Screen-reader friendly output (no spinners/panels)
`--observe`	`-O`	choice	Enable observability (`langfuse`, `langextract`); env `PRAISONAI_OBSERVE`
`--version`	`-V`	flag	Show version and exit

Flag	Type	Description
`--framework`	choice	Framework adapter to use. Choices are discovered from the adapter registry, so built-in adapters (`ag2`, `autogen`, `autogen_v4`, `crewai`, `praisonai`) and any third-party adapter installed via the `praisonai.framework_adapters` entry-point group are all valid. See Framework Adapter Plugins.
`--ui`	choice	UI: chainlit/gradio
`--auto`	remainder	Auto-generate agents
`--init`	remainder	Initialize `agents.yaml` from a task description. Prints provider-setup guidance and exits if no LLM provider is configured (see `praisonai setup`).
`--deploy`	flag	Deploy application
`--schedule`	str	Schedule pattern
`--schedule-config`	str	Schedule configuration file
`--provider`	str	Cloud provider
`--max-retries`	int	Max retry attempts
`--llm`	str	LLM model
`--model`	str	Model name
`--hf`	str	HuggingFace model
`--ollama`	str	Ollama model
`--dataset`	str	Dataset path
`--tools`	str	Tools path/names
`--no-tools`	flag	Disable tools
`--tool-retry-attempts`	int	Tool retry max attempts (default: 3)
`--tool-retry-delay`	int	Tool retry initial delay in ms (default: 1000)
`--tool-retry-backoff`	float	Tool retry backoff factor (default: 2.0)
`--tool-retry-on`	str	Tool retry error types (CSV, default: “timeout,rate_limit,connection_error”)
`--verbose`	flag	Verbose output
`--save`	flag	Save output
`--memory`	flag	Enable memory
`--user-id`	str	User ID for memory
`--planning`	flag	Planning mode
`--planning-tools`	str	Planning tools
`--planning-reasoning`	flag	Planning with reasoning
`--auto-approve-plan`	flag	Auto-approve plans
`--web-search`	flag	Native web search
`--web-fetch`	flag	Web fetch
`--prompt-caching`	flag	Prompt caching
`--max-tokens`	int	Max output tokens
`--final-agent`	str	Final agent name
`--guardrail`	str	Output validation
`--metrics`	flag	Token/cost metrics
`--telemetry`	flag	Usage monitoring
`--mcp`	str	MCP server command
`--fast-context`	str	Codebase search
`--handoff`	str	Agent delegation
`--auto-memory`	flag	Auto memory extraction
`--claude-memory`	flag	Claude memory format
`--todo`	flag	Todo generation
`--router`	flag	Smart model selection
`--trust`	flag	Auto-approve tools
`--approve-level`	str	Risk level approval
`--sandbox`	str	Sandbox mode
`--external-agent`	str	External CLI tool — choices are dynamically populated from the registry (built-ins claude/gemini/codex/cursor plus any entry-point plugins), like `--cli-backend`. Uses manager-Agent delegation by default. See Register a custom external agent
`--external-agent-direct`	flag	Use external agent as direct proxy (skip manager Agent delegation)
`--image`	str	Image analysis
`--image-generate`	flag	Image generation
`--file`	str	Input file
`--url`	str	Input URL
`--goal`	str	Goal/objective
`--auto-analyze`	flag	Auto-analyze context
`--query-rewrite`	flag	Query rewriting
`--rewrite-tools`	str	Query rewrite tools — path to tools.py, OR comma-separated names resolved via the full tool resolver (local → wrapper → SDK builtins → praisonai-tools → plugins)
`--expand-prompt`	flag	Prompt expansion
`--expand-tools`	str	Prompt expansion tools — path to tools.py, OR comma-separated names resolved via the full tool resolver (local → wrapper → SDK builtins → praisonai-tools → plugins)
`--public`	flag	Public deployment
`--merge`	flag	Merge workflows
`--claudecode`	flag	Claude Code integration
`--realtime`	flag	Realtime mode
`--call`	flag	Call mode
`--workflow`	str	Workflow file
`--workflow-var`	str	Workflow variables
`--auto-save`	str	Auto-save name
`--history`	int	History size
`--include-rules`	str	Include rules
`--no-rules`	flag	Disable auto-injection of project instruction files (AGENTS.md, CLAUDE.md, etc.)
`--append-system-prompt`	str	Append text (literal or `@file`) to the system prompt for this invocation only (`code` / `chat` / `run`). Env fallback: `PRAISONAI_APPEND_SYSTEM_PROMPT`. Never persisted. See Append System Prompt.
`--checkpoint`	str	Checkpoint ID
`--thinking`	str	Thinking budget
`--compaction`	str	Compaction strategy
`--output-style`	str	Output style
`--policy`	str	Policy file
`--background`	flag	Background execution
`--lite`	flag	Lite mode (minimal dependencies)
`praisonai chat` / `-i`	flag	Interactive TUI mode (resident on `praisonai-code`; wrapper adds richer dispatch)
`praisonai chat`	flag	Single prompt chat mode (resident on `praisonai-code`; wrapper adds richer dispatch)

As of PR #3818, chat and code open their resident interactive TUI on a standalone pip install praisonai-code — no wrapper required. Default run "…" and every run --output … mode also route through the in-process Agent (PR #2853). The wrapper (pip install praisonai) adds the richer legacy dispatch for code, gateway, and channel bots. See the standalone-limits table.

Fourteen stub command groups — agents, workflow, registry, memory, skills, hooks, rules, eval, package, templates, todo, research, commit, and call — delegate to the wrapper for their legacy entry points. On a standalone install those entry points fail fast with a single-line install hint (<group> requires the full wrapper. Install the full wrapper: pip install praisonai) and exit 1 as of PR #2854 — no more Rich ImportError traceback. Native subcommands stay standalone-safe: memory learn, skills bundle/check/eligible, eval list-judges, eval list. See the mixed-group table.

Note on --tools: Comma-separated tool names (e.g., --tools tavily_search,my_tool) are now resolved via the unified ToolResolver, so any tool reachable from YAML is also reachable from the CLI.

SDK Module Reference

praisonaiagents (Core SDK)

Module	Location	Features	CLI Exposure
Agent	`agent/agent.py`	Agent, ImageAgent, ContextAgent, DeepResearchAgent, QueryRewriterAgent, PromptExpanderAgent	Via wrapper CLI
Agents	`agents/agents.py`	Multi-agent orchestration	Via wrapper CLI
Task	`task/task.py`	Task definition	Via wrapper CLI
Tools	`tools/`	80+ tools (file, web, db, search, etc.)	`praisonai tools`
Memory	`memory/`	FileMemory, Memory, RulesManager, AutoMemory, WorkflowManager, HooksManager, DocsManager, MCPConfigManager	`praisonai memory/rules/workflow/hooks/docs/mcp`
Knowledge	`knowledge/`	RAG, chunking, vector stores, rerankers	`praisonai knowledge`
Workflows	`workflows/`	Workflow, Pipeline, Route, Parallel, Loop, Repeat	`praisonai workflow`
MCP	`mcp/`	MCP client, server, transports (HTTP, WebSocket, SSE)	`praisonai mcp`
DB	`db/`	DbAdapter protocol, lazy backends	Via wrapper
Observability	`obs/`	16 providers (Langfuse, LangSmith, AgentOps, etc.)	`--telemetry`
Eval	`eval/`	AccuracyEvaluator, PerformanceEvaluator, ReliabilityEvaluator, CriteriaEvaluator	`praisonai eval`
Skills	`skills/`	SkillManager, SkillLoader, SkillValidator	`praisonai skills`
Planning	`planning/`	Plan, PlanStep, TodoList, PlanStorage, PlanningAgent	`--planning`
Telemetry	`telemetry/`	MinimalTelemetry, TelemetryCollector, PerformanceMonitor	`--telemetry`
Guardrails	`guardrails/`	GuardrailResult, LLMGuardrail	`--guardrail`
Handoff	`agent/handoff.py`	Agent-to-agent delegation	`--handoff`
Checkpoints	`checkpoints/`	Shadow git checkpointing	`praisonai checkpoint`
Thinking	`thinking/`	Thinking budget management	`praisonai thinking`
Compaction	`compaction/`	Context compaction	`praisonai compaction`
Background	`background/`	Background task execution	Via wrapper
Hooks	`hooks/`	Event hooks, middleware	`praisonai hooks`
UI	`ui/`	AGUI, A2A	`praisonai a2a`
LLM	`llm/`	LLM client, model router, rate limiter	Internal

praisonai (Wrapper/CLI)

Module	Location	Features	CLI Exposure
CLI Main	`cli/main.py`	PraisonAI class, argparse dispatcher	`praisonai`
CLI Features	`cli/features/`	50+ feature handlers	Various commands
Integrations	`integrations/`	Built-in: Claude Code, Gemini CLI, Codex CLI, Cursor CLI — plus any registered `praisonai.external_agents` plugins	`--external-agent`
Adapters	`adapters/`	Readers, rerankers, retrievers, vector stores	Internal
Capabilities	`capabilities/`	27 LiteLLM-parity endpoints	`praisonai <capability>`
Deploy	`deploy/`	Docker, cloud providers	`praisonai deploy`
Auto	`auto.py`	AutoGenerator, WorkflowAutoGenerator	`--auto`, `workflow auto` ¹
Train	`train.py`	Model training	`praisonai train`
Scheduler	`scheduler/`	Job scheduling	`praisonai schedule`
Templates	`templates/`	Agent templates	`praisonai templates`
UI	`ui/`	Chainlit, Gradio interfaces	`praisonai ui/chat/code`

¹ workflow auto previously raised NameError on _models_cache in every code path; fixed in PR #2147.

Quick Reference

Common Commands

# Run agent with prompt
praisonai "Create a blog post about AI"

# Run workflow
praisonai workflow.yaml

# Interactive mode
praisonai chat

# Chat UI
praisonai chat

# Health checks
praisonai doctor

# Memory management
praisonai memory show                # 10 most recent
praisonai memory list --limit 25     # alias for show
praisonai memory add "Important context"

# Tool management
praisonai tools list
praisonai tools list --source registered

# Workflow management
praisonai workflow list
praisonai workflow auto "Research AI trends"

# Visual workflow builder
praisonai flow
praisonai flow import workflow.yaml --open
praisonai flow export <flow_id> -o my_flow.yaml

# Deployment
praisonai deploy init
praisonai deploy run

Common Flag Combinations

# Agent with memory and planning
praisonai "Task" --memory --planning

# Agent with web search and tools
praisonai "Research topic" --web-search --tools

# Agent with external CLI tool (delegated via manager)
praisonai "Refactor code" --external-agent claude

# Agent with external CLI tool (direct proxy)
praisonai "Refactor code" --external-agent claude --external-agent-direct

# Agent with guardrails and metrics
praisonai "Generate content" --guardrail --metrics

# CI mode with JSON output
praisonai doctor ci --json --output report.json

Error Handling

The CLI provides clean error handling across all display modes.

Budget Exceeded Errors

When agents exceed their budget limits, the CLI catches BudgetExceededError and provides actionable guidance:

# Before: Raw Python traceback
Traceback (most recent call last):
  ...
BudgetExceededError: Agent 'Researcher' exceeded budget: $1.2500 >= $1.0000

# After: Clean CLI message
Budget limit exceeded: [budget] Agent 'Researcher' exceeded budget: $1.2500 >= $1.0000
- Set max_budget parameter (e.g., Agent(max_budget=1.00))

Error Handling Features:

Works across all display modes: silent (-qq), quiet (-q), default, verbose (-v), debug (-vv), --output jsonl, --output json, --output editor
Returns exit code 1 for budget exceeded errors
Includes remediation hints in error messages
No raw Python tracebacks in production

`praisonai init`

praisonai init scaffolds a .praisonai/ project structure (config.yaml, starter agent, starter command). The scaffolded model is set to match whichever provider credential is detected in your environment:

Credential detected	Scaffolded model
`OPENAI_API_KEY`	`gpt-4o-mini`
`ANTHROPIC_API_KEY`	`anthropic/claude-3-5-sonnet-latest`
`GEMINI_API_KEY`	`gemini/gemini-1.5-flash`
`GOOGLE_API_KEY`	`google/gemini-1.5-flash`
`GROQ_API_KEY`	`groq/llama-3.3-70b-versatile`
`COHERE_API_KEY`	`cohere/command-r`
`OLLAMA_HOST`	`ollama/llama3.2`
(none detected)	`gpt-4o-mini` (fallback)

When no provider credential is detected, praisonai init prints:

No provider credential detected — scaffolded placeholder model 'gpt-4o-mini'.
Set one of OPENAI_API_KEY, ANTHROPIC_API_KEY, GEMINI_API_KEY/GOOGLE_API_KEY,
GROQ_API_KEY, COHERE_API_KEY or OLLAMA_HOST, then update the model if needed.

When a credential is detected, it prints:

Detected provider — scaffolded model: anthropic/claude-3-5-sonnet-latest

Flag	Description
`--global`	Scaffold `~/.praisonai/` instead of the project root
`--force` / `-f`	Overwrite existing files

`praisonai session resume`

praisonai session resume <id> restores full conversational state — chat history, model, and agent name — for a prior session. See the dedicated Session Resume page for full details.

# List sessions to find an ID
praisonai session list

# Resume (restores model + history)
praisonai session resume <session_id>

# Resume and continue with a new prompt
praisonai session resume <session_id> "what should I do next?"

# Show transcript only (no state restore)
praisonai session resume <session_id> --transcript

`praisonai usage`

praisonai usage reports aggregate token and cost totals from the local session store — no network, no configuration. See the dedicated Usage CLI page for options and JSON output.

# Last 30 days, grouped by day
praisonai usage

# Group by model or project
praisonai usage --by model
praisonai usage --by project

# Window and JSON output
praisonai usage --days 7
praisonai usage --json

​First-run Credential Check

​Command Tree

​Direct Prompt Examples

​Global Flags (70+ flags)

​SDK Module Reference

​praisonaiagents (Core SDK)

​praisonai (Wrapper/CLI)

​Quick Reference

​Common Commands

​Common Flag Combinations

​Error Handling

​Budget Exceeded Errors

​praisonai init

​praisonai session resume

​praisonai usage

​See Also

First-run Credential Check

Command Tree

Direct Prompt Examples

Global Flags (70+ flags)

SDK Module Reference

praisonaiagents (Core SDK)

praisonai (Wrapper/CLI)

Quick Reference

Common Commands

Common Flag Combinations

Error Handling

Budget Exceeded Errors

`praisonai init`

`praisonai session resume`

`praisonai usage`

See Also