/access · Invitation & scoping intake

Choose how you start with Forward.

Access to Agentic OS (Forward HIVE) and the Grid API is provisioned by invitation so we can confirm your model catalog, regional routing, and rate card before production traffic starts.

Startup & builder paths

Three onboarding tracks.

01 / Tracks

Select the track that matches your team. We review the requested outcome, deployment requirements and commercial terms with your team.

Agentic OS · Forward HIVE API

Request Agentic OS & Grid API access

For product and engineering teams building multi-step agent threads, isolated tool sandboxes, and low-latency Grid inference.

· Organization and workspace provisioning by invitation
· Confirmation of account-enabled agents and model catalog
· Thread, tool-access and approval configuration for the agreed scope
· Written ACU allocation, limits and effective rate confirmation

Managed Agent Pods & Automatic Implementation

Scope a Managed Agent Pod or Project

For enterprise teams that want Forward to wire tools, evaluators, human-in-the-loop approval gates, and agreed operating and support terms.

· Workflow boundary, tool surface, and acceptance metric review
· Dedicated Agent Pod or end-to-end Automatic Implementation scope
· Citation and provenance audit configuration for regulated outputs
· Written proposal separating implementation milestones from recurring ACUs

Post-Training · Agent RL & Rollouts

Scope an Agent RL or Rollout pilot

For ML teams adapting checkpoints on multi-step tool trajectories or reserving high-concurrency rollout sampling.

· Checkpoint compatibility, reward readiness and learning method review
· Deterministic verifier and reward rubric contract
· Reserved rollout sampling fleet or full-loop policy optimization quote
· Promotion gate to Grid Dedicated or Private serving

How accounts are metered once enabled.

Full pricing & ACU reference

Agent Compute Unit (`hive_acu`) + scoped platform commitment

Agentic OS (Forward HIVE) & Managed Agents

Meter multi-step agent orchestration, tool execution, code sandboxes, browser sessions, and sub-agent spawning through a single predictable unit of account.

Deterministic ACU conversion ratios live in platform metering; the account agreement confirms monetary rates, included usage and effective date.

Rollout sampling window, training tokens, or scoped pilot milestone

Agent RL, Own-Trainer Rollouts & Post-Training

Optimize open-weight checkpoints on verifiable multi-step trajectories—either by pointing your own RL trainer at Forward Grid rollout workers or co-engineering an Agent RL loop.

Quoted per checkpoint, environment complexity, and rollout concurrency window.

USD per 1M tokens (input / cached input / output) or per-check detector rate

Serverless Inference, Batch & OpenDetector

Call open-weight reasoning, coding, and multimodal models directly through an OpenAI-compatible API, or run asynchronous batch workloads at a 50% token discount on eligible models.

Published reference rates for 19 models; account-effective billing requires an approved plan version and invoice verification.

Reserved GPU / node / rack commitment over an agreed term

Dedicated Capacity & Private Deployments

Reserve single-tenant GPU capacity or isolated private deployments when throughput predictability, custom checkpoints, or strict data residency govern your workload.

Quoted per hardware configuration, exact region, and delivery window.

Need security, residency, or procurement documents first?

Review our deployment boundaries, AWS regional posture, and enterprise review checklist before submitting your workload scope.

Review Trust & Deployment