Runtime vs Hyperagent

Hyperagent and Runtime at a glance

Hyperagent is a horizontal agent platform. It began inside Airtable and was spun out as an independent company in 2026, led by Airtable co-founder Howie Liu. You build teams of specialist agents, each with its own identity, instructions, tools, skills, and memory, that delegate to each other, run on schedules, watch for changes in Live Mode, and are called from Slack, Telegram, email, webhooks, or MCP clients. Every thread gets a private computer where the agent writes and runs code, and a browser for websites. Outputs such as documents, tables, webpages, and slides land in a searchable library. Use cases span marketing, research, recruiting, operations, software, and business intelligence. It is for teams that want a flexible, self-serve place to put agents to work on almost anything.

Runtime is the AI agent harness for payment and fintech teams. Payment ops, risk, compliance, finance, underwriting, onboarding, and support build agents from their SOPs. Each agent works on its own isolated computer in your cloud, reaches your ledger, processor, and bank portals through APIs, databases, CLIs, MCP servers, and a browser, and stops for a person before anything moves money. It is for operations leaders at payment companies, banks, and lenders who need many agents running on regulated data.

Both are platforms for many agents. The difference is depth in one domain versus breadth across all of them.

How they differ

Where agents run and who holds the data

Hyperagent is hosted. Its documentation describes a private computer per thread and browser sessions on Browserbase, and does not publish a self-hosted or bring-your-own-cloud option.

Runtime runs agent computers in your AWS, GCP, or Azure account, or you self-host the whole platform with Helm, and you can bring your own sandbox provider. Credentials are masked, and card numbers and SSNs are stripped from prompts and logs. For a team with PCI scope or a sponsor bank reviewing its controls, where the agent's computer lives is usually the first question.

Models

Hyperagent has one of the widest hosted model catalogs available, from Anthropic, OpenAI, and Google to DeepSeek, Kimi, Qwen, and GLM, and lets you connect a ChatGPT subscription. That is a real strength for cost and quality tuning.

Runtime takes the next step for regulated data: bring any model, including open-weight models served inside your own cloud, so PCI and PII work never leaves your environment. Harness routing (Claude Code, Codex, OpenCode) and model routing come with fallbacks when a provider goes down.

Guardrails for money movement

Hyperagent's controls are thoughtful: an Ask first mode, approval cards that preview exact writes, and a thread that holds the full transcript of each run. Its own docs note that Ask first is a behavioral preference, not a sandbox.

Runtime's guardrails are built around payments: agents start read-only, approvals sit in front of releasing a payout, applying a reserve, booking a ledger entry, or replying to a sponsor bank, and RBAC limits who can approve what. Every run is recorded end to end, including the trigger, each query and tool call, approvals, cost, and result, and is exportable for a sponsor bank, auditor, or examiner.

Fintech domain depth and who gets you to production

Hyperagent is self-serve and general purpose by design. You learn the platform and decide what to build.

Runtime pairs you with a forward-deployed AI engineer, not an account executive plus a solutions engineer. The team built payment and fintech infrastructure at Hulu's payments team at Disney, Finix, Modern Treasury, and BlackRock's AI quant group, including risk and underwriting systems and payment and post-trade reconciliation. The FDE maps your processes, builds the first agents with your team, sets up guardrails, and trains admins. Your people use the software every day and keep using it after the FDE leaves.

One record across teams

Both platforms share memory across agents. Runtime's is organized around the transaction. A single stuck payment touches four teams: support gets the ticket, payment ops traces it, finance sees a recon break, and risk or compliance may review it. On Runtime it is one investigation with one record, and each run improves the organization's memory of resolved tickets, fraud caught, and reconciliation edge cases. When a process is solved, agents can turn it into a deterministic script that lives in your repos.

Where Hyperagent is stronger

• Breadth. Marketing, research, recruiting, software, and BI, not just operations.
• Self-serve speed. Sign up and have agents working the same day.
• Model catalog. A wide range of hosted frontier and open models in one picker.
• Rich outputs. Live webpages, slides, documents, dashboards, and media in a shared library.
• Transparent pricing. Public credit plans with per-run usage breakdowns and budget caps.

Pricing

Hyperagent publishes credit-based plans from Hyperagent 20 ($20 per month) to Hyperagent 10000 ($10,000 per month). Your plan price becomes credit plus a bonus of up to 45 percent on larger plans, agents spend it on models, browser sessions, search, and tools, and you can top up or turn on auto-recharge.

Runtime has a Free tier ($0, one session), Teams from $99 per seat per month, and Enterprise with custom pricing and self-hosting. Teams usually weigh it against the work of analysts they were about to hire. Rain replaced $250k in vendor spend with Runtime.

Which should you choose

Choose Hyperagent if

• You want agents across many functions, not one regulated domain
• Self-serve setup and pay-as-you-go credits fit how you buy
• A hosted platform is acceptable for the data involved
• Wide model choice from hosted providers is enough

Choose Runtime if

• Agents work card data, PII, ledgers, or payouts
• Agent computers and models must run in your cloud, or fully self-hosted
• You need approvals before money moves and a record your sponsor bank will accept
• You want one investigation across support, payment ops, finance, and risk
• You want an engineer who has built payment infrastructure to get the first agents live

Frequently asked questions

What is Hyperagent?

Hyperagent is an AI agent platform that started inside Airtable and was spun out as an independent company in 2026, led by Airtable co-founder Howie Liu. Agents have persistent identities, memories, and skills, work on a private computer for each thread, and can be called from Slack, Telegram, email, webhooks, MCP clients, or schedules.

Can Hyperagent run in my own cloud?

Hyperagent's public documentation describes a hosted platform and does not publish a self-hosted or bring-your-own-cloud option. Runtime runs agent computers in your AWS, GCP, or Azure account, or fully self-hosted, with your own sandbox provider if you prefer.

Which models does Hyperagent support?

Hyperagent offers a wide hosted catalog, including models from Anthropic, OpenAI, Google, xAI, DeepSeek, Moonshot, Alibaba, and others, and lets you connect a ChatGPT subscription. Runtime lets you bring any model, including open-weight models served in your own cloud for PCI and PII work.

How much does Hyperagent cost?

Hyperagent publishes monthly credit plans from $20 to $10,000 per month. Your plan price becomes credit, with a bonus of up to 45 percent on larger plans, and agents spend it on models, browser sessions, and tools. Runtime has a free tier, Teams from $99 per seat per month, and custom Enterprise pricing.