# ifivo

> The runtime control plane for AI agents. See what agents do, approve what matters, stop them instantly — with audit-grade logs across every tool, API, and payment rail. Works with ChatGPT, Claude, and Gemini via MCP.

ifivo sits in front of the tools your AI agents use (Stripe, AWS, ad platforms, OpenAI, and any HTTP API). Agents call the Agent Gateway with their intended action; ifivo enforces deterministic policies (block-wins precedence, AND semantics within a policy), routes high-risk actions to a human approval queue, and exposes an instant org-wide kill switch. Every action is logged with identity, intent, risk score, and decision — immutably.

## Integrate in one minute

- MCP endpoint (ChatGPT / Claude / Gemini): `POST https://www.ifivo.com/api/mcp` with header `Authorization: Bearer ifv_YOUR_ORG_KEY`
- Agent Gateway (from your agents): `POST https://www.ifivo.com/api/gateway/actions` with body `{"agent_api_key":"sk_live_…","vendor":"stripe","action":"refund","amount_cents":22000,"metadata":{...}}`
- Health check: `GET https://www.ifivo.com/api/health`
- Get your org API key: sign in at https://www.ifivo.com/login then open Settings

## Docs

- [Integration quick-start](https://www.ifivo.com/integrate): Copy-paste configs for ChatGPT, Claude, Gemini, plus curl examples and MCP tool schemas.
- [MCP setup guide](https://www.ifivo.com/docs/mcp): Full guide with assistant-by-assistant configuration JSON and example prompts.
- [llms-full.txt](https://www.ifivo.com/llms-full.txt): Verbose machine-readable reference with every endpoint, every MCP tool, request/response shapes, and policy DSL.

## MCP tools exposed

- `list_agents` — each agent's status, owner, daily budget, spend today, risk score
- `list_transactions` — filter by status (allowed / pending_approval / blocked / killed / approved / denied) and agent
- `list_pending_approvals` — what's waiting for a human decision right now
- `approve_transaction` / `deny_transaction` — resolve a pending approval by id
- `engage_org_kill_switch` / `release_org_kill_switch` — stop or resume all agent traffic for the org
- `quarantine_agent` — freeze a single agent without affecting the rest
- `create_policy` — define a deterministic rule (vendor, action, amount, risk_score, time_of_day, etc.)
- `summarize_today` — natural-language recap of the day's agent activity

## Live demo

- Home: https://www.ifivo.com
- App (requires sign-in, demo tenant is pre-seeded with 247 transactions, 5 agents, 39 pending approvals, $40K spend): https://www.ifivo.com/app

## Pricing

- Starter: Free. 1 org, 5 agents, 10k actions/month, shadow mode + dashboards.
- Team: $499/mo. Unlimited agents, 500k actions/month, approvals + kill switch, SSO, audit export.
- Enterprise: Custom. SOC 2 artifacts, on-prem gateway option, dedicated support.

## Contact

- GitHub: https://github.com/marcosgcunha/ifivo
- Sign in: https://www.ifivo.com/login