# Spikefrost

> Spikefrost is the AI operating system for enterprises. A company's own team builds agentic apps in weeks with their own Claude Code or Codex; Spikefrost supplies the tools (connectors, data, deployment, every AI model from one team balance), the rules (isolation, permissions, approval gates, spending limits, audit) and the method the agents follow.

Spikefrost is operated by Spikefrost Limited, a company registered in England and Wales (Company Number 16978632), based at 71-75 Shelton Street, Covent Garden, London, WC2H 9JQ, United Kingdom.

## What Spikefrost does
- Give a company one place for all of its AI: create a team, add funds to the team balance, and every member can use any AI model and build agentic apps from that balance.
- Govern spend and access per member: roles with daily and monthly spending limits, department budgets, cost alerts, and per-turn cost attribution to the member, app and job.
- Offer every major model family from the one balance: Claude (Anthropic), GPT (OpenAI), Gemini (Google), Kimi (Moonshot AI), DeepSeek, Qwen (Alibaba), Llama (Meta), GLM (Z.ai), Mistral, plus image and video generation, at public usage rates (https://c.spikefrost.com/pricing).
- Build, govern, and run production AI agents and agentic apps on a secure runtime.
- Run every app in its own V8 isolate, with a separate database, key-value store, object state, and metrics dataset per app AND per environment. This is isolate-level isolation on shared infrastructure, not dedicated single-tenant compute.
- Apply granular, identity-based RBAC scoped down to individual tools, enforced server-side.
- Mint short-lived, per-call credentials so long-lived keys never touch agent code.
- Provide full audit logging and observability: every action, job, and cost is attributable and replayable.
- Connect securely to Slack, Teams, email, Stripe, webhooks, and any MCP server, owned and scoped per agent.
- Deploy globally to a stateless edge runtime via the sf CLI.

## Key facts
- Category: AI operating system for enterprises (enterprise SaaS, AI agent platform, developer platform).
- Security posture: SOC 2 Type II, ISO 27001, GDPR data processing agreement available, isolate-level per-app isolation, encryption in transit and at rest.
- Availability: 99.9% monthly uptime SLA for app serving, with service credits. Runs on Cloudflare Workers, whose own 99.99% SLO covers the compute path and excludes storage bindings.
- Recovery: 30-day point-in-time database recovery; full commit and deployment audit history.
- Enterprise capabilities (provisioned under an enterprise agreement, not enabled on standard plans): a dedicated Cloudflare account operated by Spikefrost, a customer-owned Cloudflare account (requires the customer's own Workers for Platforms entitlement), managed Postgres alongside D1 for datasets over 10 GB or Postgres-specific needs, and residency for stored data via a region-selectable Postgres database.
- Not offered on any tier: residency for COMPUTE (code runs in the location nearest the user), dedicated single-tenant hardware, customer-managed encryption keys, on-premise or air-gapped deployment, private network/VPC peering, SSO/SAML/SCIM for platform sign-in, HIPAA.
- Data policy: customer data is never used to train AI models.
- Primary users: security-first and regulated engineering teams.
- Plans: Free (3 apps, 1 seat, 1 environment, 24-hour logs), Pro (15 apps, 3 seats, 2 environments, 7-day logs), Business (50 apps, 10 seats, 5 environments, 30-day logs), Enterprise (custom). Runtime and AI usage is pay-as-you-go from a prepaid balance.
- Prerequisite: building with Spikefrost requires one active Claude, Codex, Kimi or Grok coding-agent subscription, separate from the Spikefrost plan.
- Install: desktop app for macOS and Windows at https://www.spikefrost.com/en/download, or the sf CLI via `curl -fsSL https://edge.spikefrost.com/sf/install.sh | sh` (macOS, Linux, WSL), then `sf login`.
- Full page index for LLMs: https://www.spikefrost.com/llms-full.txt

## Pages
- [Home](https://www.spikefrost.com/en/): What Spikefrost is, how a team, its balance and every model fit together, and what an agentic app is.
- [Pricing](https://www.spikefrost.com/en/pricing): Plans (Free, Pro $25, Business $99 per team per month, Enterprise custom), what the team balance pays for, and the coding-subscription prerequisite.
- [Download](https://www.spikefrost.com/en/download): Spikefrost desktop app for macOS and Windows, plus the one-line sf CLI installer.
- [Getting started](https://www.spikefrost.com/en/docs/getting-started): Install Spikefrost and Claude Code or Codex, create a first app, build by chat, deploy.
- [Set up Claude Code or Codex](https://www.spikefrost.com/en/setup-claude): Step-by-step install and sign-in for the coding agent Spikefrost uses to build apps.
- [Docs home](https://www.spikefrost.com/en/docs): Platform definition, certifications, SLA, isolation, audit and recovery in one place.
- [Changelog](https://www.spikefrost.com/en/changelog): Dated release notes for the platform, CLI and desktop app.

## Docs
- [Platform architecture](https://www.spikefrost.com/en/docs/platform-architecture): How requests, apps, agents, data stores and the edge runtime fit together.
- [Isolation and tenancy](https://www.spikefrost.com/en/docs/isolation-and-tenancy): What V8 isolates guarantee and what they do not; per-app and per-environment isolation for a security review.
- [SLAs, limits and quotas](https://www.spikefrost.com/en/docs/sla-and-limits): 99.9% uptime commitment, support response times, certifications, and platform limits.
- [Observability and audit](https://www.spikefrost.com/en/docs/observability-and-audit): Every signal recorded, retention periods, and the audit trail of AI actions.
- [Failure domains and resilience](https://www.spikefrost.com/en/docs/failure-domains): Where the SLA boundary sits and what fails independently.
- [Durability and recovery](https://www.spikefrost.com/en/docs/durability-and-recovery): 30-day point-in-time database recovery and one-minute rollback.
- [Enterprise capabilities](https://www.spikefrost.com/en/docs/enterprise-capabilities): Dedicated or customer-owned Cloudflare account, Postgres alongside D1, data residency; what is and is not on the enterprise tier.
- [Deployment models](https://www.spikefrost.com/en/docs/deployment-models): Shared infrastructure, dedicated account operated by Spikefrost, or a customer-owned account.
- [How agents work](https://www.spikefrost.com/en/docs/agents-overview): Agents are TypeScript classes deployed inside your app's own worker.
- [Writing an agent](https://www.spikefrost.com/en/docs/writing-an-agent): The agent class, tools, model selection and deployment.
- [Human approvals](https://www.spikefrost.com/en/docs/human-approvals): Gating agent actions behind a person's approval.
- [Channels overview](https://www.spikefrost.com/en/docs/channels-overview): Slack, Teams, email, web chat, messengers and webhooks; how an agent claims a connector.
- [Choosing a data store](https://www.spikefrost.com/en/docs/data-stores): D1, Postgres, Durable Objects, KV and files compared.
- [Environments](https://www.spikefrost.com/en/docs/environments): Staging and dev inside the same app, branch mapping and promotion.
- [The sf CLI](https://www.spikefrost.com/en/docs/sf-cli): Install, authenticate, build, deploy, data, logs and secrets from the terminal.
- [Desktop app](https://www.spikefrost.com/en/docs/desktop-app): What the macOS and Windows app does and how it relates to the CLI.
- [Backoffice run by agents](https://www.spikefrost.com/en/docs/backoffice-agents): Ten operational jobs an agent can hold, and the rules that keep handover safe.

## Guides
- [Enterprise AI agents: the complete guide](https://www.spikefrost.com/en/learn/enterprise-ai-agents): What enterprise AI agents are, how they are built and governed, and how to ship them safely.
- [AI customer support agents](https://www.spikefrost.com/en/learn/ai-customer-support-agents): Resolve tickets across Slack, email and chat, escalating to humans when needed.
- [AI sales agents](https://www.spikefrost.com/en/learn/ai-sales-agents): Qualify leads, follow up, and keep the CRM current across channels.
- [AI marketing agents](https://www.spikefrost.com/en/learn/ai-marketing-agents): Draft and schedule campaigns, run promotions and report on performance.
- [AI agents for operations](https://www.spikefrost.com/en/learn/ai-agents-for-operations): Reconcile data, watch metrics on a schedule, run checks and report.
- [AI agents for ecommerce](https://www.spikefrost.com/en/learn/ai-agents-for-ecommerce): Promotions toward a sales goal, buyer questions and anomaly flags.
- [Ecommerce automation guide](https://www.spikefrost.com/en/learn/ecommerce-automation): How growing brands automate operations and ship features faster.

## Comparisons
- [Spikefrost vs open-source agent frameworks](https://www.spikefrost.com/en/compare/spikefrost-vs-agent-frameworks): LangChain, CrewAI and AutoGen are libraries you host yourself; Spikefrost is a managed, secure runtime.
- [Spikefrost vs automation tools](https://www.spikefrost.com/en/compare/spikefrost-vs-automation-tools): Zapier, n8n and Make run fixed workflows; Spikefrost runs agents that decide each step toward a goal.
- [Spikefrost vs building in-house](https://www.spikefrost.com/en/compare/spikefrost-vs-building-in-house): Build or buy an enterprise agent platform: time, cost and risk trade-offs.

## Glossary
- [What is an AI agent?](https://www.spikefrost.com/en/glossary/what-is-an-ai-agent): An AI system given a goal, tools and the authority to act, choosing each step rather than following a script.
- [What is an agentic app?](https://www.spikefrost.com/en/glossary/what-is-an-agentic-app): Software that operates itself: a team of agents running a real process end to end.
- [AI agent vs workflow](https://www.spikefrost.com/en/glossary/ai-agent-vs-workflow): A workflow follows a fixed sequence; an agent decides what to do at each step.
- [What is MCP (Model Context Protocol)?](https://www.spikefrost.com/en/glossary/model-context-protocol-mcp): The open standard for connecting agents to tools and data; Spikefrost connects to any MCP server.

## Solutions
- [Customer support](https://www.spikefrost.com/en/solutions/customer-support): Deploy a support agent that resolves tickets across Slack, email and chat.
- [Sales](https://www.spikefrost.com/en/solutions/sales): A sales agent that qualifies leads, follows up, books meetings and updates the CRM.
- [Ecommerce](https://www.spikefrost.com/en/solutions/ecommerce): Automate promotions, support and ops, and ship store features the same day.
- [Operations](https://www.spikefrost.com/en/solutions/operations): Reconciliation, monitoring, scheduled reporting and internal Q&A, auditable by design.

## Optional
- [Blog](https://www.spikefrost.com/en/blog): Posts on guardrails, credentials, SOC 2, multi-agent systems and ROI; the full list with summaries is in llms-full.txt.
- [Privacy Policy](https://www.spikefrost.com/en/privacy): How Spikefrost collects, uses, stores, and protects personal data under UK GDPR.
- [Terms of Service](https://www.spikefrost.com/en/terms): Terms governing access to and use of the Spikefrost platform.

## Contact
- General: support@spikefrost.com
- Privacy: privacy@spikefrost.com
- Security: security@spikefrost.com
- Website: https://www.spikefrost.com