Back to GBG for Agents
    Beta

    The trust layer for the agents behind your customer relationships.

    Have confidence in every agentic decision, powered by GBG Go.

    Your AI agents now run the customer journeys that matter, from targeting and onboarding new customers to re-verifying existing ones. Every one of those moments relies on a trust operation, and an agent left to guess puts it at risk.

    GoPlane is the identity verification layer for AI agents: an MCP server and CLI that run every identity, KYC or fraud check as a full GBG Go journey rather than a one off call, so each decision becomes a journey run you can replay months later rather than a model guess. It is in Beta today, and your Go key is your GoPlane key.

    Consumer
    AI Agent
    Onboard
    actions
    GoPlane
    Trust control plane
    GBG Go
    Deterministic journey
    outcomes
    The catalogue

    Everything GBG Go can do, now native to your agents.

    GBG Go is one platform with 110+ identity, fraud and risk capabilities across 195+ countries. Rather than wrapping a slice of that, GoPlane hands your agents the whole catalogue, along with the journeys built on top of it, as things they can actually run.

    Identity data

    Verify a person against national, credit and telco data in 195+ countries.

    Documents and biometrics

    Act on documents and biometrics captured with our existing SDKs.

    Fraud and risk

    Score against a cross-industry network of 1,000+ brands and spot serial fraudsters.

    KYC and compliance

    Run screening, re-verification and step-up checks against local regulation.

    What GoPlane adds

    Why not just call the Go API?

    Your agent can already call the Go API and get an answer back. GoPlane turns that call into a journey run, so each journey your agent runs gets a run id you can look up later and every step it took stays on the record. A run that needs more information stays open and continues when the missing piece arrives, and results come back in a fixed shape with exit codes your code can check.

    There is a second version of that question: why can’t my agent handle this on its own? The identity, fraud and risk data behind these decisions sits under entitlements a model cannot reach, and reading it the way a risk team would is domain work it has to guess at. GoPlane puts that execution layer underneath the agent so it can do the work properly.

    AI agent alone vs AI agent + GoPlane

    Onboard a customer
    AI agent alone
    Runs the steps it thinks it needs, no guarantee the required ones ran
    AI agent + GoPlane
    Runs every check in the journey, each one on the record
    Handle a refer
    AI agent alone
    No way to pick it back up
    AI agent + GoPlane
    Branches, holds the run open, resumes it when the gap is filled
    Repeat a decision
    AI agent alone
    No guarantee of the same answer twice
    AI agent + GoPlane
    Same input, same outcome, every time
    Show its working
    AI agent alone
    Hard to trace
    AI agent + GoPlane
    Every step tied to a run id you can retrieve
    Work at 3am
    AI agent alone
    Yes, unverifiable
    AI agent + GoPlane
    Yes, and you can check it in the morning
    Where teams start

    What teams use it for first.

    Most teams don’t start with new customer onboarding. They start with the work sitting behind it: referred cases waiting on a human, the KYC refresh backlog, the re-verification nobody has capacity for. Runs that can wait and continue are built for exactly that kind of work, and the same journeys carry through to onboarding when you’re ready.

    Referred cases

    A referred case sits open until someone picks it up. An agent can gather what is missing, continue the same run, and hand a complete file to a human reviewer.

    High-volume onboarding

    Sellers, drivers, couriers, players. A lot of applicants, a lot of documents, and a lot of people who don't finish first time. The run stays open until they come back.

    KYC refresh and remediation

    Work through a backlog unattended, using an answers file for anything the journey asks for. Every decision recorded.

    Step-up checks

    Send someone to a document or biometric check only when the risk score calls for it, and show afterwards why you did.

    See it run

    Watch an agent onboard a customer

    GoPlane runs as a CLI and an MCP server the agent calls, so a run normally happens without anyone watching. Here it is made visible: pick an applicant and start onboarding. On the left is the agent reasoning, on the right is each GBG Go journey step it triggers and what comes back.

    Applicant
    GoalOnboard Mia, or refer if it isn't clear-cut
    Score identityRefer to review
    Agent - reasoning loopplans · observes · decides
    Pick an applicant and run onboarding.
    GoPlane - Grounded in GBG GoGBG Go journey runs
    Each step here is an actual journey run.

    Sample data, illustrative flows for demo purposes.

    For developers

    One engine. CLI or MCP.

    GoPlane ships as a single static binary with two interfaces over one shared execution engine. A journey behaves identically whether a human on a laptop or an AI agent over MCP is driving it.

    For humans & CI
    CLI

    Type commands, pipe JSON, branch on exit codes. Ships as one static binary. Drop it in a Dockerfile or Makefile and go.

    $ goplane journey run 5f2a…c9 \
        --input applicant.json \
        -o json --watch
    For AI agents
    MCP server

    Run goplane mcp and an agent calls the same capabilities as tools over stdio. No bespoke wrapper, no drift between what humans and agents can do.

    $ goplane mcp
    # stdio server ready
    # tools: run_journey, get_run_status, …

    The CLI is an agent API too. An agent can shell out to goplane … -o json and branch on exit codes. Often leaner than wiring MCP end-to-end.

    Capabilities

    What GoPlane does today.

    Run and monitor existing GBG Go journeys (identity checks, document and address verification, and GBG Trust fraud scoring) with structured output and stable exit codes.

    Run journeys

    Execute a published journey through an agent.

    Monitor & fetch state

    Poll a run to a terminal status, or snapshot its current state without waiting.

    Interactive runs

    A journey that needs more info waits. Supply it and resume the same run. Same id, same audit trail.

    Autonomous mode

    Run unattended to completion, filling interaction gaps from a supplied answers file.

    List & get journeys

    Discover published journeys and fetch a journey definition.

    MCP server

    The same capabilities exposed as tools an AI agent calls over stdio.

    Structured output

    --output json|yaml, wrapped in { version, kind, data, warnings } for scripts and agents.

    Stable exit codes

    Deterministic exit codes so CI pipelines and agents can branch reliably.

    Coming soon

    Prompt-based journeys.

    Describe the journey you want. The AI Journey Builder generates a draft you can validate, publish and run through the same lifecycle you already know.

    goplane - prompt-based journey
    # describe the journey
    $ goplane journey create --prompt "EMEA identity + fraud journey, no document check"
    → drafting via AI Journey Builder…
    → draft: j_draft_9a2b
    $ goplane journey validate j_draft_9a2b
    → ok · 0 errors · 2 warnings
    $ goplane journey publish j_draft_9a2b
    → hash: 5f2a…c9 · ready to run
    goplane - foresight optimise
    # foresight scans recent runs
    $ goplane foresight analyse journey_emea_identity
    → 12,481 runs · 7d window
    → step "doc_check": 38% redundant · suggest gate on risk_score > 0.4
    → projected: −22% latency · −17% cost · no lift in false-accept
    $ goplane foresight apply --as-draft
    → draft: j_draft_7c1e · review before publish
    In Go today, coming to agents

    Foresight-driven continuous optimization.

    Foresight already runs in GBG Go. It watches live journey runs, finds steps that aren’t earning their place, and suggests journeys that cost less and run faster without moving the decision boundary. What’s coming is agent access, so your agents can call it directly and act on what it finds.

    195+
    Countries covered
    110+
    Go capabilities
    1,000+
    Brands in the fraud network
    2
    Interfaces (CLI, MCP)
    8
    GoPlane capabilities
    Getting started

    Your Go key is your GoPlane key.

    GoPlane uses the GBG Go API key you already have. No second set of credentials and nothing to put through procurement. Self-serve signup comes at GA. Until then we set Beta users up ourselves.

    Existing GBG Go customers

    Your Go key already works. Tell us which journeys you want your agents to run and we’ll get you set up.

    Talk to us
    New to GBG Go

    GoPlane runs on Go, so we’ll set up both together. Start with the platform, then add the agent layer.

    Explore GBG Go

    Important notice·GBG is not responsible for any decisions, whether automated or not, that the Customer makes based on the information, Service, Results or data provided (including, but not limited to onboarding decisions) through any GBG for Agents capability.

    Talk to us about GoPlane.

    GoPlane is in Beta. Tell us what you’re building and we’ll get you set up.

    We'll only use your details to talk to you about GoPlane.

    FAQ

    Frequently asked questions

    What is GoPlane?

    GoPlane is the identity verification layer for AI agents, and the trust layer for the agents behind your customer relationships. It is an MCP server and CLI that run every identity, KYC or fraud check through GBG Go as a real journey, so the same inputs produce the same decision and every step ties back to a run id you can look up later.

    Why not just call the GBG Go API from my agent?

    You can, and you'll get your answer. What you won't get is a record of the whole journey, a run that can wait for missing information and continue, or results in a shape your code can rely on. GoPlane adds those. Later you can pull up the run and show what your agent decided and why.

    Why can't my agent handle this on its own?

    The identity, fraud and risk data behind these decisions sits under entitlements a model cannot reach, and reading it the way a risk team would is domain work it has to guess at. GoPlane puts that execution layer underneath the agent, so the agent decides what to do and GoPlane runs the check properly. Compare the two side by side at https://agents.gbg.com/goplane.html#agent-alone-vs-goplane

    Who is GoPlane for?

    Teams who already have AI agents doing real work, and now want those agents to handle identity, KYC and fraud checks without losing the record of what happened. In practice that means two people: whoever is building the agents, and whoever in risk has to sign off on what those agents decide.

    What do teams use GoPlane for first?

    Most teams start with the work sitting behind new customer onboarding: referred cases waiting on a human, the KYC refresh and remediation backlog, step-up checks, and high-volume onboarding where applicants do not finish first time. Runs that can wait and continue suit that work, and the same journeys carry through to onboarding later.

    Do I need a GBG Go account?

    Yes. GoPlane runs on GBG Go and uses your existing Go API key. There's no separate GoPlane credential. If you're not a Go customer yet, we'll set up both together.

    How do I get access during Beta?

    Email us on this page and tell us what you're building. We set Beta users up directly. Self-serve signup comes at GA.

    Is GoPlane available?

    Yes. GoPlane is available now, in Beta. Your Go key is your GoPlane key, so there's nothing extra to buy. Email us on this page and we'll set you up.

    What is GBG Go?

    GBG Go is our trust intelligence platform. It puts 110+ identity, fraud and risk capabilities on one global platform, with a visual journey builder for the onboarding, compliance and risk journeys businesses run every day.

    What is grounding?

    Grounding anchors an AI agent's decisions to verified, retrievable source material instead of the model's internal guess. It's the mechanism that stops hallucinations and makes multi-step actions safe to run. That's what GoPlane does for identity checks: every one runs through GBG Go, so identical inputs always produce the identical decision, and every step ties back to a run you can retrieve.

    What is an MCP?

    MCP stands for Model Context Protocol. It's a shared standard that lets AI assistants like Claude, Cursor, Copilot or ChatGPT plug into outside tools safely. Think of it as a universal socket: once a product exposes an MCP server, any MCP-compatible AI can use its features without a custom integration. GoPlane exposes its capabilities over MCP so agents can run real GBG Go journeys as tools.

    What is a CLI?

    CLI stands for Command Line Interface. It's a small program you run in a terminal by typing commands, instead of clicking around a website. Developers, scripts and AI agents can all call a CLI to do work and get results back. GoPlane ships as a single CLI binary, and it behaves the same way as the MCP server, so a build pipeline and an agent get identical results.

    What is an agent experience layer?

    GoPlane is the agent experience layer on top of GBG Go. It gives your agents every trust operation Go can run, whether that's verifying an identity, running KYC, assessing risk or stopping fraud, and it does it natively. No wrestling with the API, no bespoke integration.

    What decisions does GoPlane return?

    Every identity check returns one of three clear outcomes (accept, refer, or reject) with the weak signal named. Accept onboards, refer hands off to a human reviewer, and reject escalates to your fraud team. Every step ties back to a run id you can retrieve later.