Repeatable team workflows
Turn a documented process such as weekly reporting, request triage or lead qualification into a shared agent.
Independent tool overview
Workspace Agents at a glance
ChatGPT Workspace Agents are Codex-powered team agents that run repeatable, multi-step workflows using approved apps, tools, files and organizational controls.
Visit the official Workspace Agents site ↗
- Availability
- Business, Enterprise and Edu
- Builder
- Codex-powered
- Channels
- ChatGPT, Slack and API
- Schedules
- Supported
- Integrations
- Apps, tools and custom MCPs
- Billing
- Token-based credits
Overview
What Workspace Agents is
Workspace Agents move beyond a personal chatbot by packaging a repeatable process into an agent that a team can share. A builder describes the job, adds instructions and tools, tests the workflow and publishes a controlled version to the organization.
Agents run in the cloud and can use files, skills, apps, custom MCP connections and memory. They can be launched in ChatGPT, run on a schedule, triggered through an API or deployed to Slack channels, subject to plan, workspace and administrator settings.
The most important implementation work is governance, not prompting. Teams need scoped connections, test cases, approval rules, named owners and monitoring—especially when an agent can send messages, update records or use an agent-owned account on behalf of many people.
Use cases
Who Workspace Agents is best for
The strongest fit depends on the job you need the product to complete, not the size of its feature list.
Cross-tool operations
Gather context from approved systems, transform it and deliver an output without manual copy-and-paste.
Slack-based assistance
Make a governed agent available in a team channel for questions, scheduled updates or supported actions.
Capabilities
Core Workspace Agents features
Guided agent builder
Uses ChatGPT to define instructions, steps, models, reasoning effort and starter prompts.
Connected apps and tools
Adds enabled apps, custom MCPs, files, skills, web search and other permitted capabilities.
Shared publishing
Keeps an agent private, shares it by link or publishes it to the organization's directory.
Multiple trigger channels
Runs agents interactively, on schedules, through API triggers or in connected Slack channels.
Connection controls
Supports end-user connections or agent-owned shared connections plus action and parameter constraints.
Administrative governance
Provides role-based controls, agent activity visibility and the ability to suspend agents where supported.
Process
How the Workspace Agents workflow works
Step 1
Choose a controlled process
Start with a frequent, measurable workflow that has clear inputs, outputs and an accountable owner.
Step 2
Configure least-privilege access
Connect only required apps and limit repositories, actions, parameters and shared credentials.
Step 3
Test before publishing
Use normal, adversarial and failure examples to check output quality and prevent unsafe actions.
Step 4
Publish and monitor
Release to the smallest necessary audience, inspect usage and revise or suspend the agent when behavior drifts.
Cost
Workspace Agents pricing and free plan
Workspace Agents require an eligible ChatGPT workspace. Runs use the same agentic usage and credit pool as supported Codex and ChatGPT Work activity. OpenAI's current Business and Enterprise/Edu rate card prices runs by tokens rather than a fixed amount; a typical GPT-5.6 run is estimated at 5–25 credits.
ChatGPT Business
Plan fee + included usage
Eligible Business seats include Workspace Agents and draw initial activity from the shared agentic usage allowance.
- Business starts at two users
- Included limits apply
- Flexible credits can extend eligible usage
- Check current seat pricing at purchase
Flexible credit usage
Token-based
GPT-5.6 agent runs currently consume credits by input, cached input and output tokens.
- 100 credits per 1M input tokens
- 10 credits per 1M cached input tokens
- 500 credits per 1M output tokens
- Typical run estimate: 5–25 credits
Enterprise and Edu
Contract pricing
Workspace access, included usage and billing terms depend on the organization's agreement.
- Admin enablement and RBAC supported
- Flexible or token-based terms may apply
- Confirm the applicable rate card with OpenAI
Pricing checked . Check current pricing at the source ↗
Assessment
Workspace Agents strengths and limitations
Where it stands out
- Packages a team process once for repeated use across the organization
- Can keep running in the cloud without an open user session
- Works across ChatGPT, schedules, API triggers and supported Slack channels
- Combines instructions, files, memory, skills and connected business systems
- Role-based publishing and access controls support managed rollout
- Preview and draft publishing reduce the chance of changing a live agent accidentally
What to consider
- Not available on personal ChatGPT plans
- Usage cost varies with context, task complexity, model and output length
- Agent outputs and actions can be inaccurate even when the workflow appears repeatable
- Slack deployment requires a paid Slack plan, administrator setup and shared app connections
- Using a personal account as an agent-owned connection can expose that account's authority to other users
- Parameter constraints narrow action inputs but do not filter the data an app returns
- Some building, analytics, sharing and deployment tasks still require Agent Studio rather than the Codex plugin
- High-impact or irreversible actions need explicit approval, audit logs and a manual recovery path
Compare
Workspace Agents alternatives
The right alternative depends on the specific output, workflow, controls and budget your project requires.
Agents
Zapier Agents
A cross-app agent platform centered on Zapier's large automation ecosystem.
Explore Zapier Agents →Business Operations
Gumloop
A visual no-code builder for AI workflows with explicit nodes and data movement.
Explore Gumloop →Business Operations
Microsoft Copilot
A Microsoft-centered assistant option for organizations standardized on Microsoft 365 and Copilot tooling.
Explore Microsoft Copilot →Questions
Workspace Agents FAQs
What are ChatGPT Workspace Agents?
They are shared, Codex-powered agents for repeatable organizational workflows. They can use approved tools and run in ChatGPT, Slack, on a schedule or through an API trigger.
Who can use Workspace Agents?
OpenAI lists Workspace Agents for eligible ChatGPT Business, Enterprise and Edu workspaces. Availability and building permissions depend on administrator settings and roles.
How much does a Workspace Agent run cost?
There is no fixed per-run charge. Current credit use depends on model tokens and complexity; OpenAI estimates a typical end-to-end GPT-5.6 run at roughly 5–25 credits.
Can Workspace Agents work in Slack?
Yes. The organization must complete ChatGPT and Slack admin setup, use a paid Slack plan and configure agent-owned shared connections for the agent's apps.
Can an agent take actions in company systems?
It can use supported app actions when those apps, credentials and permissions are explicitly configured. Teams should constrain access and require confirmation for consequential changes.
Is Codex running the agent when I use the Workspace Agents plugin?
No. In the Codex plugin, Codex acts as a builder that edits and publishes the configuration. Running and previewing the agent still happens through supported Workspace Agent surfaces.
Bottom line
Our Workspace Agents verdict
Workspace Agents are strongest when a team already has a stable process and wants one governed implementation across ChatGPT and Slack. They are not a shortcut around process design: connection scope, evaluations, approvals, ownership and cost monitoring determine whether the automation is safe and economical.
Visit Workspace Agents website ↗Similar Agents Tools
View all Agents AI tools →
Perplexity Personal Computer
Perplexity Personal Computer - Perplexity's local orchestrator working across your files, native apps, and the web
Gemini Enterprise Agent Platform
Gemini Enterprise Agent Platform - Google's successor to Vertex AI and platform for building, scaling, and governing enterprise agents across 200+ models
Holo 3
Holo 3 - H Company's open-weight computer-use agent that navigates enterprise desktops with SOTA accuracy on 10B active params
Grok Voice Think Fast 1.0
Grok Voice Think Fast 1.0 - xAI's new voice agent with background reasoning and tool calling for complex support workflows
AI training for the future of work.
Get access to all our AI courses, hundreds of real-world AI use cases, live expert-led workshops, an exclusive network of AI early adopters, and more.
AI Courses
Get unlimited access to all of our current & upcoming industry-specific AI courses for the duration of your subscription.
Daily Guides
To keep up with the rapid pace of AI, our team publishes AI implementation guides daily. Our library contains 300+ practical use cases to automate real-world work.
Workshops
Join weekly, live, interactive sessions with industry leaders who are at the forefront of AI for hands-on implementation guidance and exclusive insights.
Community
Network with an exclusive community of AI-first professionals who are working smarter with AI. Learn how early adopters are using AI in their work and businesses.