Meet Surogate.

AI colleagues for every team. Design, test, govern and run them without code, in your cloud or on your own hardware.

Start buildingContact Sales

The payables agent checks a supplier statement against policy, then asks before it pays. Filmed on a demo business.

Give each agent a job.

Six markets, six agents, each filmed at work on a demo business.

Checks a supplier statement against the payables policy, then holds the €12,400 payment until a person says yes.

Van Dijk Freight, a demo business

By market

Brief it like a colleague.

Say what the agent is for. The Copilot shows a plan, builds only what you approve, and anything that reaches people or money needs its own yes.

From a blank page to a live agent.

Six steps, each one a screen in the studio.

  1. 01

    Set it up without code

    Every part of an agent is a screen in the studio. Change one and the next session uses it, with no rollout.

    • Start from a template or a blank agent, and talk to it straight away
    • The Copilot can take any step, and builds only what you approve
    Supplier deskSaved, next session uses it
    • IdentityWho the agent is, how it speaks, what it refuses.
    • SkillsProcedures in markdown or drawn on the canvas.
    • Knowledge BasesYour documents as a searchable wiki.
    • ToolsMCP servers and Composio toolkits, scanned first.
    • ModelOurs, your fine-tunes or any OpenAI-compatible endpoint.
    • CapabilitiesBrowser, Claude Code and Codex, research, schedules.
    • CodingRepositories it can branch and open pull requests on.
    • GovernanceAllowed tools, approvals and browsing rules.

  2. 02

    Teach it your work

    Attach skills and knowledge bases. Your documents become a searchable wiki, with the source passage behind every answer.

    • Every skill version is kept, and each session records the one it used
    • MCP servers and Composio toolkits, scanned before the agent sees them
    A policy handbook attached to the payables agent from the library.

  3. 03

    Put it where your team works

    A person who moves from the web to Slack or WhatsApp keeps the same conversations, memory and allowance.

    Web app
    Website widget
    API
    Slack
    Microsoft Teams
    WhatsApp
    Telegram
    Discord
    Phone

    One agent. Same conversations, memory and allowance on each.

  4. 04

    Let it run for days

    Goals, missions, loops and deep research run for hours, days or months. Programs follow up with people on the days you choose.

    • Sub-agents share a board of checked notes, so a dead end is not repeated
    • A task is stored and gets up to 3 attempts, each reading what went wrong before
    /goalWorks until a separate evaluator says the goal is met.
    /missionSplits an objective into tasks for sub-agents.
    /loopRuns a prompt every few minutes, hours or days.
    /deep-researchReturns a cited report.
    Day 0EnrolledDay 1WelcomeDay 3Check-inDay 7ReminderDay 8Team alerted

  5. 05

    Keep it inside the rules

    Sensitive calls wait in the approvals inbox. Rules are fixed for each turn, so text injected mid-turn cannot loosen them.

    • Keys the agent never sees: the proxy adds credentials on the way out
    • Sessions survive crashes, and failed steps are undone in reverse order
    A payment over the limit waits in the inbox until a person approves it.

  6. 06

    Make it better from its own sessions

    Mark sessions as good or bad, and the studio turns them into training data, fine-tunes, scores and experts.

    • Skill optimisation keeps the best version, and nothing changes until you apply it
    • Train on AWS, Azure, GCP, Nebius, RunPod, Modal, Oracle, Vast.ai or your own servers
    1. SessionsMarked good or bad
    2. DatasetsFrom sessions or synthetic
    3. TrainingSFT, DPO, GRPO, distillation
    4. EvaluationsBenchmarks and A/B runs
    5. ExpertsFine-tuned specialists

Sell an agent you built.

Put a price on it and share a link. Buyers do not need a Surogate account.

  • Your Stripe account

    Subscriptions, usage packs or both, in seven currencies. You are the merchant of record.

  • Pages made for you

    A buy page at your agent's address, and a landing page you assemble from blocks.

  • Packages

    Each offer opens the channels, tools, knowledge and model tier you choose.

Run all of it inside your own network.

On-premise or air-gapped, the whole platform runs on your Kubernetes with your models, and no data has to leave your network.

Your network

  • Studio and runtime

    Agents, sessions and schedules, on your Kubernetes

  • Sandboxes

    Shell, browser and office files, a network policy per session

  • Models

    Surogate 3.7, Jackrabbit and Amami on your GPUs, next to your fine-tunes

  • Vault

    API keys and passwords, added by a proxy, never shown

  • Storage

    Sessions, files and logs; skills, datasets and models

  • Your servers

    Train and serve over SSH, or on your own cloud account

The runtime is open source: github.com/invergent-ai/surogates

Why teams pick it.

Live in minutes, on every channel your people use, and open for your security team to read.

every part of an agent is a screen in the studio, from setup to approvals
No code
every part of an agent is a screen in the studio, from setup to approvals
for one agent: web, Slack, Teams, WhatsApp, the phone and more
9 channels
for one agent: web, Slack, Teams, WhatsApp, the phone and more
runtime on GitHub, so your security team can read every line
Open source
runtime on GitHub, so your security team can read every line

How it compares

Agent frameworks

LangChain, LangGraph

Elsewhere: You write the agent in code, then build sessions, sandboxes, channels and governance around it.

In Surogate: Those come with the platform, and agents are built in the studio.

Coding agents

Claude Code, Codex

Elsewhere: One person, on one machine, in one terminal.

In Surogate: Whole teams on nine channels, with governance and a record of every step. Claude Code and Codex can run inside a session.

Managed agent platforms

Elsewhere: The vendor builds and runs the agents on its own cloud.

In Surogate: Your team builds them, in your cloud or air-gapped, with any model.

One plan per client organisation.

Unlimited users and agents on every plan. Models at the provider's price, or bring your own.

  • Pilot

    $0for the first 60 days

    • Up to 1,000 agents working at once
    • 40 hours web browsing
    • 50 GB workspace, 200 GB hub storage
    • 40 GPU-hours for fine-tuning and evals

  • Production

    $990a month

    • Up to 1,000 agents working at once
    • 200 hours web browsing
    • 250 GB workspace, 1 TB hub storage
    • 40 GPU-hours for fine-tuning and evals

  • Scale

    $2,990a month

    • Up to 5,000 agents working at once
    • 600 hours web browsing
    • 750 GB workspace, 3 TB hub storage
    • 200 GPU-hours for fine-tuning and evals

Personal plans start free with five agents and no card.

See all plans

Questions

Do I need to write code?

No. Every part of an agent is a screen in the studio: identity, skills, knowledge, tools, model, capabilities and governance. The Copilot can build any step for you, and builds only what you approve.

Which models can an agent use?

Ours (Surogate 3.7, Jackrabbit and Amami), your own fine-tunes, OpenRouter or any OpenAI-compatible endpoint.

Can it run on our own hardware?

Yes. The whole platform runs on your Kubernetes, on-premise or air-gapped, with your models, and no data has to leave your network.

How do approvals work?

You choose the calls that need approval. They wait in the approvals inbox, and a grant covers one call, for one hour. Anything that reaches people or money needs its own yes.

Is it open source?

The runtime is, on GitHub, so your security team can read every line.

Can I sell an agent I built?

Yes. Put a price on it and share a link. Buyers pay on your own Stripe account and do not need a Surogate account.

What does it cost?

Each client organisation starts with a 60-day pilot at $0, then Production at $990 a month or Scale at $2,990 a month, with unlimited users and agents. Personal plans start free.

Meet your first agent.

Start free with five agents and no card, or install the platform on your own Kubernetes.

Start building