需要什么帮助？

Get started with Frontal

Learn the building blocks of Frontal and choose the shortest path to a useful production system.

Quickstart

Make your first authenticated call and read a result back, in about five minutes.

Your first agent

Give an agent a brief, a tool, and a source, then read the trace it leaves.

Your first workflow

Build a durable workflow with a deterministic step, an agent decision, and an approval before an external action.

Get started with Frontal

Learn the building blocks of Frontal and choose the shortest path to a useful production system.

Quickstart

Make your first authenticated call and read a result back, in about five minutes.

Installation

Install the SDK or the CLI, and set the environment variables both expect.

Authentication

Authenticate people, applications, and CI jobs without widening access beyond the work they need to perform.

Workspaces

How workspaces isolate data, people, and billing, and when to create a second one.

Your first agent

Give an agent a brief, a tool, and a source, then read the trace it leaves.

Your first workflow

Build a durable workflow with a deterministic step, an agent decision, and an approval before an external action.

Connect a data source

Authorise a source, choose its scope, and map it onto the ontology.

Model your domain

Define the smallest shared model that gives your first agent or workflow reliable business context.

Build an agent

From a blank brief to an agent you would let a colleague use.

Build a workflow

Turn a repeatable business process into a durable, inspectable sequence of steps.

Add an approval step

Put human judgement in front of consequential actions while preserving the context needed to make the decision.

Evaluate a change

Tell whether a new version is better before it reaches anyone.

Debug a run

Use the run trace to find whether a failure came from context, a tool, a policy, or execution.

Deploy to production

Move from a workspace you experiment in to one people depend on.

Integrations overview

Connect Frontal to the places work begins, the systems that receive actions, and the tools teams already use.

Model Context Protocol

Use Frontal through MCP clients while preserving the same workspace, permission, and action boundaries.

Webhooks

Start work from external events and deliver completed outcomes to systems that need them.

Export runs and evidence

Send operational data to the systems where teams analyse, retain, and act on it.

Connectors overview

Bring enterprise records into Frontal as governed, traceable context for agents and workflows.

Connector lifecycle

Understand what happens from authorization through synchronization, maintenance, and disconnection.

Syncs and permissions

Keep source access narrow, explain what a sync read, and make data failures visible before agents use the results.

Frontal CLI

Inspect and automate Frontal from a terminal or CI job using the same workspace controls as the API.

CLI automation and CI

Use scoped credentials, structured output, and explicit failure handling when automating Frontal from CI.

Frontal Agent

Give people a conversational operating surface for business context, guided investigation, and approved actions.

Conversations and actions

Turn a question into a traceable answer or a governed request for work.

Build agents

Create bounded agents that reason over business context, use governed tools, and leave an inspectable record of every run.

Tools and context

Give an agent the smallest set of trusted information and actions that can complete its job.

Memory and limits

Keep useful state across work without creating unbounded access, cost, or autonomous behaviour.

Agent runs and versions

Evaluate an agent before promotion, inspect its execution after launch, and roll back without losing evidence.

Workflows

Orchestrate durable business processes that combine deterministic steps, agent judgement, waits, approvals, and recovery.

Workflow authoring

Design a workflow as explicit triggers, state transitions, and ownership boundaries.

Approvals and recovery

Pause for human decisions, recover from failed steps, and avoid duplicated external actions.

Data

Turn enterprise source records into governed, traceable context that agents and workflows can use.

Quality and lineage

Make source, transformation, validation, and freshness part of every decision that uses data.

Storage and retrieval

Preserve artifacts, documents, indexes, and durable state as part of the context an intelligent system can use.

Ontology

Give people, agents, and workflows one shared model of the entities and relationships that make up your business.

Entities and relationships

Define the business objects and typed links that make a question answerable across systems.

Source mapping and versions

Map changing source systems into stable business definitions and promote ontology changes safely.

Evaluation

Measure whether agents, workflows, and intelligent applications meet the requirements you set before and after release.

Case sets and promotions

Build representative cases, compare versions, and make promotion decisions from evidence rather than intuition.

Observability

Make every intelligent-system run explainable through traces, metrics, costs, and evidence trails.

Traces and alerts

Investigate a single execution in detail and alert on changes that require attention.

Security and governance

Control who can access context, which actions may run, and how every consequential decision can be reviewed.

Access and audit

Apply roles, attributes, key scopes, policies, and approvals at the point where work is performed.

Data controls

Apply isolation, encryption, residency, retention, and deletion policies to the context that powers your systems.

Frontal API

Integrate Frontal programmatically through versioned HTTP resources for agents, workflows, context, and operations.

Resources and lifecycle

Understand the public objects that make up a Frontal workspace and the lifecycle events they produce.

Deployment

Choose an operating model for Frontal and promote intelligent systems through controlled environments.

Production readiness

Prepare an agent or workflow for real users with clear ownership, controls, recovery, and evidence.

Errors

What each failure means and which ones are worth retrying.

Rate limits

The limits, the headers that report them, and how to stay under them.

Lifecycle statuses

Interpret the states shared by long-running synchronization, workflow, agent, and evaluation work.