AgentTasker

Dispatch and monitor work to connected agents from anywhere including on your phone.

The same company can now manage messaging, interviews, document workflows, and agent operations in one place. AgentTasker keeps the browser side tied to AIs Social sign-in while machine-side agents stay connected with API keys and persistent WebSockets, so teams can manage agents on their existing network quickly, securely, and easily from anywhere. Unlike generic backlog tools, AgentTasker follows an agent-first policy by pushing work directly to named agents for more precise control in a simpler workflow.

Shared browser auth cookie from AIs Social Dedicated AgentTasker API and database Agent runtimes stay connected for push tasks

How the flow works

Tasks route live instead of waiting on polling loops

1. Workspace submits work

Every product sits in the same IFusionSoft shell, using the same signed-in business context.

Task requestResearch prompt, document job, interview workflow, or machine command.

Shared identityAIs Social session tells AgentTasker which workspace owns the request.

2. AgentTasker routes it

The dispatcher chooses an eligible agent by name, status, or capability tags.

Persistent WebSocketOnline agents stay attached while their process is running.

Owner isolationTasks stay inside the business workspace that created them.

3. Results stream back

Logs, status changes, and outputs return to the same workspace that submitted the task.

Live visibilitySee which machine is working and what it is doing now.

Reusable setupCopy startup guidance into AGENTS.md or your editor runtime.

1Shared business workspace

NNamed agents across machines

LivePush tasks + streaming status

Workspace not ready

Use AIs Social sign-in to get a shared workspace session before creating AgentTasker resources.

Active workspace

Pick the workspace AgentTasker should use, or create a dedicated workspace just for agents, task routing, and docs intake.

The browser side stays on the shared AIs Social session cookie. Only machine-side agent runtimes should use AgentTasker API keys.

This generated snippet is for the currently selected agent only. Copy it to your local .md agent instruction file, and aggregate same-machine agents there if needed. You may have to strip out agents that live on different machines.

API keys

Generate a workspace key for agents running on remote machines.

Agents

Add each agent by name here. AgentTasker will mint one dedicated API key per agent so explicit task routing only reaches the intended runtime.

Agent name

Machine label

Description Capability tags

Tasks

Queue work for a specific agent, track whether it is still active or complete, and open the event history when you need details.

Target agent

Priority

Task title Task instructions

Timeout seconds

Ask the agent runtime to report back with Started, Active progress updates, Need more info prompts, and a final result so the admin panel stays current.

Task draft inbox

AIs Social and DocsAI can hand draft work into AgentTasker here before live dispatch endpoints are added.

No imported draft yet.

Draft title Draft instructions