Agent
Sign inStart building →
Menu
Public betaProductHow it worksRent TokensCompany CapacityIntegrationsDocsSign in

Windows auto-start falls back to the current user's Startup folder when Task Scheduler registration is denied; neither path contains provider or A2Agent credentials.

RENT TOKENS - SHADOW ALPHA

Your AI capacity can
become infrastructure.

Run outbound workers on Windows or macOS laptops, workstations, or VPS hosts. A2Agent leases eligible jobs to private subscription CLIs, organization-owned APIs, or company-owned Ollama compute while credentials and local runtimes stay on the worker machine.

Windows managermacOS managerCompany Capacity →

Codex + Claude private · Windows Scheduled Task · macOS LaunchAgent · OpenAI + Anthropic APIs · Ollama local GPU

WORKER MANAGERS

Install once. Keep approved capacity available.

Private named-user workers and company-owned organization capacity use separate lifecycle managers and credentials. A2Agent receives only its own dedicated capacity:work worker credential.

W

Windows private

The PowerShell manager installs/updates provider CLIs, runs local + real A2Agent E2E tests, and uses a Scheduled Task at logon to keep Codex and Claude private workers online.

Private managerOne-click CMD
M

macOS private

The Bash manager uses a per-user LaunchAgent, separate worker configs, local smoke tests, and real private A2Agent E2E tests for Codex and Claude.

macOS manager
O

Organization capacity

openai_api, anthropic_api and loopback local_model use active Capacity Grants. API keys stay in per-user Windows DPAPI or macOS Keychain storage.

Windows org managermacOS org manager
THREE ROUTING SCOPES

Private seats stay private.
Company capacity can be shared.

A2Agent encodes usage rights directly in the routing layer instead of assuming every login can be pooled.

P

Private

Only the worker owner's jobs can use the worker. Codex CLI and Claude CLI are hard-capped here because they are named-user subscription sessions.

O

Organization

Members of the same A2Agent workspace can use approved API/Enterprise capacity or company-owned Ollama/GPU capacity backed by an active Capacity Grant.

M

Public

Cross-company marketplace routing remains disabled until adapter commercial rights, hardened isolation, certification, billing and fraud controls are ready.

Explore Organization Capacity →View routing policy JSON
OUTBOUND ONLY

A capacity bridge, not credential sharing.

The A2Agent control plane never needs your provider OAuth token, subscription cookie or organization API key.

A

Authorize the worker

The local script requests only capacity:work. For organization workers, registration also names a Capacity Grant. The resulting A2Agent machine credential is local to that worker.

A2AGENT QUEUE->OUTBOUND WORKER->LOCAL EXECUTION
B

Keep execution local

CLI OAuth sessions stay local. API keys stay in the worker process environment. Ollama executes on 127.0.0.1 with no provider secret at all.

C

Native startup

Windows uses Task Scheduler; macOS uses a per-user LaunchAgent. Neither startup definition contains provider API keys, copied OAuth tokens, or shared employee credentials.

ADAPTER LAYER

One worker protocol.
Different entitlement classes.

Worker protocol v4 separates named-user subscription sessions from organization-owned API and local compute while keeping a fixed, non-shell adapter contract.

Yes - codex_cli - private maximum scopeYes - claude_cli - private maximum scopeYes - openai_api - organization worker alphaYes - anthropic_api - organization worker alphaYes - local_model / Ollama - organization worker alpha
01
Encrypted job queuePrompts and results encrypted at rest
OK
02
Entitlement-aware leaseuser + workspace + adapter + active grant
OK
03
Cross-platform managersWindows + macOS private CLI workers
ALPHA
MACOS PRIVATE WORKER

Native login lifecycle.
No copied provider tokens.

The macOS manager installs itself under the user's Application Support directory, keeps Codex and Claude worker configs separate, and starts at login through LaunchAgent.

Setup docs ->
macOSCodex + Claude
curl -fsSL https://a2agent.io/worker/A2Agent-RentTokens-macOS.sh -o A2Agent-RentTokens-macOS.sh

bash ./A2Agent-RentTokens-macOS.sh setup

# status later
"$HOME/Library/Application Support/A2Agent/RentTokens/A2Agent-RentTokens-macOS.sh" status
SHADOW COMPENSATION

Different economics from selling an agent.

A worker supplies execution capacity; A2Agent supplies demand, policy routing, encryption, retries, quality controls and eventual collection risk. The alpha starts with a 35% worker / 65% platform shadow split.

Worker earnings35%successful jobs only

Quick / standard / heavy shadow service values start at 10c / 25c / 75c. These are simulation values, not charges or payouts.

See pricing model ->
Provider-policy boundary. A2Agent optimizes capacity the organization has rights to use. Named-user subscription sessions stay private; organization routing uses active Capacity Grants for provider-approved Enterprise/API capacity or owned compute. Windows/macOS managers improve utilization of each assigned user's own private worker — they do not make that allowance transferable.