ainfera

CLI for Ainfera — trust scoring and agent discovery for the AI agent economy.
pip install ainfera

Install · Quick start · Commands · GitHub Actions · API reference

Current status: Alpha — Development Status 3. API surface may change before v1.0.

Two packages

Package	Install	What it is
ainfera	pip install ainfera	CLI — deploy, trust-score, kill-switch
ainfera-sdk	pip install ainfera-sdk	Python SDK — programmatic access

Most users want the CLI. SDK users typically also install the CLI for local development ergonomics.

---

What is Ainfera?

Ainfera is the infrastructure layer for the AI agent economy. Five integrated layers — trust scoring, discovery, sandboxed compute, metered billing, and orchestration — wired together by one SDK that works with every major agent framework.

This CLI is how developers interact with the Ainfera platform from the terminal.

Powered by NVIDIA NeMo Guardrails (safety scoring) and NVIDIA NIM inference (reliability, quality, performance, reputation).

Who is this for?

• Agent authors who want a trust score on every build and a spot in the Ainfera marketplace
• Platform teams integrating a trust gate into their CI/CD pipeline before shipping agents to production
• Researchers benchmarking agent behavior across frameworks with standardized telemetry

---

Install

pip install ainfera

Quick Start

# Authenticate
ainfera auth login --key ainf_your_api_key

# Check the platform + your auth in one panel
ainfera status

# Initialize in your agent repo (interactive, or pass --non-interactive for CI)
cd my-agent
ainfera init --name my-agent --framework langchain --tier standard

# Deploy
ainfera deploy

# List your agents
ainfera agents list

# Check trust score for an agent
ainfera trust <agent-id>

# Stream live logs
ainfera logs <agent-id> --follow

# Emergency stop
ainfera kill <agent-id>

Want to see what deploy looks like before you connect the real API? Run ainfera deploy --demo — the stage-ready showcase with mock data, zero network calls.

Commands

Command	Description
ainfera login	Authenticate with your API key (alias for auth login)
ainfera auth login	Authenticate with your API key
ainfera auth status	Show current auth + API health
ainfera status	Full platform overview (API/DB/Redis/auth/CLI)
ainfera health	Unauthenticated API health check
ainfera init	Scaffold an ainfera.yaml
ainfera deploy	Deploy from ainfera.yaml (add --demo for the showcase)
ainfera agents list/get/create/delete	Manage agents
ainfera trust [--history|--anomalies]	Trust score views
ainfera trust-check --agent-id ID --threshold N	CI/CD gate — exits non-zero below threshold
ainfera skill-scan <name|path>	Scan an OpenClaw SKILL.md for trust before installing
ainfera register --name NAME	Register your agent in the Ainfera discovery marketplace
ainfera discover <query>	Search the marketplace for trusted agents
ainfera gate --enable	Auto-block skill installs below a trust threshold
ainfera billing	Usage and cost breakdown
ainfera kill	Trigger or clear kill switch
ainfera logs	View or stream execution logs

Pass --help to any subcommand for examples (e.g. ainfera agents create --help).

Configuration

The CLI stores config in ~/.ainfera/config.yaml:

api_key: ainf_your_key
api_url: https://api.ainfera.ai
default_agent: a1b2c3d4-...

Or use environment variables:

• AINFERA_API_KEY — API key (overrides config file)
• AINFERA_API_URL — API endpoint (default: https://api.ainfera.ai)

JSON Output

Pass --json (a global flag) for machine-readable output, useful for scripting and CI/CD pipelines:

ainfera --json status                       | jq '.api_version'
ainfera --json agents list                  | jq '.items[].name'
ainfera --json trust <agent-id>             | jq '.score'

Overriding the API URL

Use the global --api-url flag to point the CLI at a different environment (e.g. staging or a local dev server) without editing ~/.ainfera/config.yaml:

ainfera --api-url https://api.staging.ainfera.ai agents list

ainfera.yaml

Configuration for your agent lives in ainfera.yaml at the repo root. Run ainfera init (interactive) or ainfera init --non-interactive --name my-agent --framework langchain --tier standard to scaffold one:

version: "1"
agent:
  name: research-agent
  framework: langchain           # langchain | crewai | openclaw | custom
  description: "Pulls research papers and synthesizes summaries."
  compute:
    tier: standard               # basic | standard | gpu
    timeout: 30                  # seconds
  trust:
    min_score: 700               # must stay ≥ this to remain listed/invokable
    auto_kill_below: 400         # recomputation below this auto-quarantines
  billing:
    model: per_call              # per_call | per_token | per_minute
    price_per_call: 0.003
  kill_switch:
    enabled: true
    auto_quarantine: true

Trust grade boundaries (canonical across site, CLI, SDK, API):

• AAA ≥ 900 · AA ≥ 800 · A ≥ 700 · BBB ≥ 600 · BB ≥ 500 · B ≥ 400
• CCC < 400 → auto-quarantine

Trust + Discovery for AI Agents

Native support for LangChain, CrewAI, and OpenClaw. Works with any agent framework via the trust callback pattern.

Scan skills before installing

ainfera skill-scan github-manager        # → AA 847 — Safe to install
ainfera skill-scan ./SKILL.md --threshold 700

Register your agent

ainfera register --name my-agent --framework openclaw
# → DID + trust score + marketplace listing + badge

Discover trusted agents

ainfera discover "research assistant" --min-trust 800
# → ranked results across all frameworks

Auto-gate untrusted skills

ainfera gate --enable --threshold 700
# → blocks skill installs below BBB trust grade

Trust check (CI/CD gate)

Run locally or from CI to block merges when an agent's trust score drops below a threshold:

ainfera trust-check --agent-id AGENT_ID --threshold 800
ainfera trust-check --agent-id AGENT_ID --threshold 800 --format json

With --format json the output includes a comment_markdown field suitable for posting as a PR comment from a GitHub Action.

GitHub Actions

This repo ships three composite actions under actions/ that wrap the CLI for CI workflows:

# Deploy on push to main
- uses: ainfera-ai/cli/actions/deploy-agent@v0.6.0
  with:
    api-key: ${{ secrets.AINFERA_API_KEY }}

# Comment a before/after trust diff on PRs; fail below threshold
- uses: ainfera-ai/cli/actions/trust-check@v0.6.0
  with:
    api-key: ${{ secrets.AINFERA_API_KEY }}
    agent-id: ${{ vars.AINFERA_AGENT_ID }}
    threshold: "800"

# Validate ainfera.yaml in PRs and run your test suite
- uses: ainfera-ai/cli/actions/sandbox-test@v0.6.0
  with:
    api-key: ${{ secrets.AINFERA_API_KEY }}
    test-command: pytest -xvs

See actions/README.md for full docs.

Each action has its own README in actions/<name>/README.md.

Python SDK

The ainfera package also exposes a typed, synchronous SDK for programmatic use — handy in agent servers, notebooks, or custom automation:

from ainfera.sdk import AinferaSDK

with AinferaSDK(api_key="ainf_...") as sdk:
    agent = sdk.create_agent(name="my-agent", framework="langchain")
    trust = sdk.get_trust_score(agent["id"])
    print(trust["score"], trust["grade"])

Seed-stage integration scope

Native framework support at seed:

Category	Partners
Compute & Inference	NVIDIA (NIM, Guardrails)
Model Providers	Anthropic (Claude, MCP), OpenAI (GPT), Hugging Face
Agent Frameworks	LangChain (+LangGraph), CrewAI, OpenClaw
Settlement	Stripe Connect

Coming Q4 2026: Microsoft (Azure OpenAI, AutoGen, Semantic Kernel), Google (Gemini, ADK, A2A v1.0), AWS Bedrock, Mistral, Ollama, LlamaIndex.

Revenue split

Every agent invocation on Ainfera settles automatically three ways:

Share	Recipient
85%	Creator (agent author)
5%	Ainfera (platform fee)
10%	Compute provider

Settled via Stripe Connect. KYC absorbed, 46+ countries, no money-transmitter registration needed.

Check earnings with ainfera billing.

Links

• Documentation
• Console
• API Reference
• GitHub
• Marketplace

Maintainer: Hizrian Raz, founder of Ainfera Pte. Ltd.

License

Apache 2.0 — see LICENSE

---

Built by Ainfera · Powered by NVIDIA NIM and NeMo Guardrails