TL;DR: OpenAI confirmed that ChatGPT workspace agents will transition to credit-based pricing on July 6, 2026, ending a 26-day free period for Business and Enterprise workspaces. The change affects agent usage across all eligible plans, requiring administrators to model costs before deployment. According to OpenAI’s release notes, the free evaluation window began June 10, 2026.
On June 10, 2026, OpenAI activated ChatGPT workspace agents for all eligible Business and Enterprise workspaces. The feature launched with a temporary free tier. That free period lasts exactly 26 days. Administrators now face a hard deadline: July 6, 2026, when credit-based pricing takes effect and every agent invocation starts consuming plan resources.
What Are ChatGPT Workspace Agents and Why Do They Matter?
ChatGPT workspace agents are persistent, configurable AI assistants deployed within an organization’s ChatGPT environment. According to OpenAI’s enterprise documentation, these agents can be created, shared, and managed directly inside ChatGPT or through Slack integration. They operate with workspace-level context, meaning they access organizational knowledge without requiring manual data uploads per conversation.
The core distinction from standard ChatGPT conversations is persistence. A workspace agent retains its instructions, connected data sources, and permission scope across sessions. Users can invoke the same agent repeatedly. The agent maintains consistent behavior. This makes them suitable for repeatable business processes like internal support ticketing, document analysis, and automated research workflows.
OpenAI’s release notes confirm that workspace agents entered general availability on June 10, 2026. The deployment covered eligible Business and Enterprise workspaces simultaneously. However, the rollout explicitly excludes one major category: Enterprise workspaces using Enterprise Key Management (EKM). OpenAI documented this limitation at launch, leaving EKM-enabled organizations without agent access until a later, unspecified date.
The 26-day free window gives organizations a brief evaluation period. After July 6, every agent interaction draws from a credit pool tied to the workspace’s plan tier.\n
How Will Workspace Agents Be Priced After July 6?
Starting July 6, 2026, ChatGPT workspace agents will consume credits based on usage, according to OpenAI’s release notes. The free evaluation period—running from June 10 to July 6—gives administrators 26 days to test agent workloads without credit deductions. Once the deadline passes, all agent activity transitions to the standard credit system already governing other OpenAI features.
OpenAI has not published a standalone rate card exclusively for workspace agents. Instead, credit consumption follows the same framework outlined in the Codex rate card, which defines how credits map to actions across Plus, Pro, Business, Enterprise, Edu, Health, and Government plans. Each plan tier receives a different credit allocation, and agent operations draw from that pool proportionally to the complexity and length of the task.
Administrators need to model costs carefully. A workspace agent that performs multi-step reasoning, retrieves documents, and generates long outputs will consume more credits than a simple single-turn query. The 26-day window exists specifically so organizations can measure actual usage patterns before the billing mechanism activates.
The pricing shift means something concrete. Workspaces that deploy agents broadly across hundreds of users could see credit pools deplete faster than anticipated. OpenAI’s documentation recommends monitoring credit usage through workspace analytics dashboards during the free period to extrapolate post-July 6 costs.
Which ChatGPT Plans Support Workspace Agents Today?
As of the June 10, 2026 launch, ChatGPT workspace agents are available for Business and Enterprise workspace tiers specifically. According to OpenAI’s workspace agents documentation, these two plan categories form the initial deployment boundary. The feature is not yet extended to Team, Plus, or Free tiers.
However, there is one critical exclusion within the Enterprise tier itself. Enterprise workspaces with Enterprise Key Management (EKM) enabled do not have access to workspace agents at launch. OpenAI’s release notes state this explicitly: agents are not available for EKM workspaces in the current release cycle. Organizations requiring EKM for compliance or regulatory reasons face a blocking dependency.
The plan eligibility matrix also extends to how credits are allocated. The Codex rate card documents separate credit structures for Edu, Health, and Government plans, but workspace agents are currently scoped to Business and Enterprise only. Edu plan administrators should not assume agent availability despite having credit allocations for other features like Codex.
For organizations evaluating deployment, the current eligibility list is narrow:
- ChatGPT Business: Full workspace agent access, credit-based after July 6
- ChatGPT Enterprise: Full access, excluding EKM-enabled workspaces
- ChatGPT Team: No workspace agent access
- ChatGPT Plus: No workspace agent access
- ChatGPT Free: No workspace agent access
- ChatGPT Edu: No workspace agent access at this time
- ChatGPT Health: No workspace agent access at this time
- ChatGPT Gov: No workspace agent access at this time
| Plan Tier | Agent Access | EKM Support | Credit Pricing |
|---|---|---|---|
| Business | Yes | N/A | July 6, 2026 |
| Enterprise (standard) | Yes | N/A | July 6, 2026 |
| Enterprise (EKM) | No | Blocked | N/A |
| Team | No | N/A | N/A |
| Edu / Health / Gov | No | N/A | N/A |
How Does Codex Integration Work With Enterprise Plans?
OpenAI’s Codex documentation describes Codex as a coding agent available across multiple ChatGPT plan tiers, including Enterprise. Codex operates as a distinct feature from workspace agents but shares the same underlying credit system. For Enterprise workspaces, Codex functions as an autonomous programming assistant capable of handling multi-file code changes, running tests, and generating pull request descriptions.
The Codex rate card breaks down credit consumption by plan type. Enterprise plans receive credit allocations that differ from Business, Plus, and Pro tiers. Each Codex action—whether it’s analyzing a repository, generating code, or executing a test suite—draws from the workspace’s shared credit pool. This means Codex usage and workspace agent usage after July 6 will compete for the same credit budget.
Enterprise administrators should understand the interaction. If a workspace deploys both Codex for engineering teams and workspace agents for operations teams, credit consumption accelerates across both verticals. The 26-day free period for workspace agents does not apply to Codex—Codex has been credit-based since its own launch.
The documentation also notes that Codex requires specific setup steps for Enterprise environments. Administrators must connect code repositories and configure access permissions before the agent can operate. This setup process is separate from workspace agent configuration, though both features are managed through the same Enterprise admin console.
What Are the Security Limitations for Enterprise Workspaces?
Enterprise workspaces face a primary security limitation: ChatGPT workspace agents are not available for environments with Enterprise Key Management enabled. OpenAI’s release notes confirm this restriction applies at launch. EKM allows organizations to hold their own encryption keys, ensuring OpenAI cannot access workspace data without the customer’s key. This feature is critical for regulated industries.
The EKM exclusion creates a direct trade-off. Organizations that require EKM for compliance—common in healthcare, finance, and government sectors—cannot deploy workspace agents. They must choose between the encryption guarantee and agent functionality. OpenAI has not announced a timeline for EKM compatibility.
According to the Enterprise admin quickstart, workspace administrators manage security through identity controls, access policies, roles, and monitoring configurations. The admin console provides oversight for workspace settings. Agent permissions fit within this framework—administrators control who can create, share, and invoke agents.
However, the workspace agents documentation notes that agents can be shared via ChatGPT and Slack. This sharing surface area introduces considerations for data governance. An agent with access to internal documents, when shared to a Slack channel, extends its data reach through that integration point. Administrators must audit agent permissions and connected data sources before enabling broad distribution.
The release notes also confirm that all apps with sync functionality now support in-region data residency. This improvement addresses a separate but related concern: where workspace data physically resides. In-region support helps organizations meet data sovereignty requirements, but it does not resolve the EKM gap for workspace agents specifically.
How Do Administrators Deploy and Manage ChatGPT Workspace Agents?
Administrators manage workspace agents through the ChatGPT admin console, where they configure identity, access, workspace settings, roles, permissions, security controls, and monitoring. According to OpenAI’s admin quickstart documentation, the recommended setup guides administrators through each configuration step to ensure proper governance before agents go live. The admin console provides centralized control.
Deployment begins with workspace configuration. Administrators define which users can create, share, and interact with agents. The system integrates with Slack, allowing agents to function across both ChatGPT and Slack interfaces. This dual-channel support means teams access agents where they already work.
Management includes several key capabilities:
- Agent creation controls: Administrators decide who can build custom agents within the workspace
- Sharing permissions: Granular settings govern whether agents remain private or circulate across teams
- Security policies: Enterprise-grade controls including data handling rules and access restrictions
- Monitoring dashboards: Usage analytics help track agent interactions and resource consumption
- Slack integration: Agents deployed through Slack inherit Enterprise workspace permissions
- EKM exclusion: Workspaces with Enterprise Key Management do not support workspace agents at launch
- Identity management: Single sign-on and SCIM provisioning integrate with existing infrastructure
- Audit logging: Administrators can review agent activity for compliance and security reviews
The EKM limitation matters. Organizations requiring customer-managed encryption keys cannot use workspace agents yet, restricting deployment for regulated industries like finance and healthcare.
| Admin Capability | Description | Availability |
|---|---|---|
| Agent Creation | Define who builds agents | All Enterprise plans |
| Sharing Controls | Manage agent distribution | Business and Enterprise |
| Security Policies | Configure data restrictions | Enterprise, Edu, Health, Gov |
| Monitoring | Track usage and costs | All workspace types |
| Slack Integration | Deploy agents in Slack | Business and Enterprise |
| EKM Support | Customer-managed encryption | Not available at launch |
Why does this centralized approach matter? Without administrative oversight, agents could expose sensitive data or consume credits unpredictably.
What Happened During the Free Workspace Agents Window?
OpenAI enabled workspace agents at no additional cost for eligible workspaces, giving organizations a window to test functionality before credit-based pricing took effect. The free period began when agents launched, with credit requirements kicking in on July 6. This gave businesses approximately 26 days to evaluate agent capabilities without budget concerns.
The free window served as both a promotion and a testing phase. Organizations could build custom agents, integrate them with Slack, and measure productivity gains. No credit consumption meant teams experimented freely.
Key aspects of the free period:
- Eligibility: Applied to Business and Enterprise workspaces without EKM enabled
- No credit deduction: Agent interactions did not consume plan credits during the window
- Full functionality: All agent features remained available without restriction
- Slack access: Agents functioned in both ChatGPT and Slack during the trial
- Transition notice: OpenAI communicated the July 6 cutoff through release notes
- Post-July 6 behavior: Agents turn off without sufficient credits
- Budget planning window: The 26-day period allowed administrators to forecast costs
- Usage data: Organizations could analyze interaction patterns to estimate future spend
The transition to paid credits changes the calculus. Teams that built workflows around free agents now face per-interaction costs. Administrators must evaluate which agents deliver sufficient value to justify ongoing credit consumption.
The release notes state that workspace agents are “off” when workspaces lack additional credits after July 6. This binary on/off behavior means partial degradation does not occur — agents either function fully or not at all.
How Does In-Region Data Sync Work for Enterprise Apps?
All apps with sync now support in-region data sync, ensuring that enterprise application data remains within designated geographic boundaries. According to OpenAI’s release notes, this capability applies to synced apps across Enterprise workspaces, addressing data residency requirements that regulated organizations face.
In-region sync means data processed through connected applications stays within the customer’s chosen region. This matters for GDPR compliance and similar regulations.
The feature operates through several mechanisms:
- Geographic routing: Synced data routes to regional infrastructure matching the workspace configuration
- Application compatibility: All apps supporting sync include in-region data handling
- Enterprise workspace requirement: The feature targets Enterprise-tier deployments
- Regulatory alignment: Supports compliance with EU, UK, and other regional data protection frameworks
- No additional configuration: In-region sync activates automatically for eligible workspaces
- Sync continuity: Existing synced apps gain in-region support without reconfiguration
- Data residency guarantees: Data remains within boundaries throughout the sync lifecycle
- Audit trail: Administrators can verify regional data handling through workspace logs
Why is in-region sync critical for enterprise adoption? Without geographic data controls, organizations in regulated sectors cannot deploy AI tools that process customer or internal data.
The expansion of in-region sync to all synced apps represents a foundational infrastructure change. Enterprise customers previously faced limitations when connecting applications that processed data across borders. The updated sync capability removes that barrier for new deployments.
What Should Businesses Budget for Agent Workloads?
Businesses must budget for agent credit consumption based on interaction volume, agent complexity, and the specific models powering each agent. Codex rate cards from OpenAI detail how credit rates apply across Plus, Pro, Business, Enterprise, Edu, Health, and Gov plans, with each plan receiving different credit allocations and consumption rates.
Budgeting requires understanding several cost factors:
- Base plan credits: Each plan tier includes a credit allocation for agent interactions
- Additional credits: Organizations purchase supplementary credits when base allocations deplete
- Agent model selection: Different models consume credits at different rates
- Interaction frequency: Teams using agents frequently exhaust credits faster
- Slack usage: Agents accessed through Slack count toward the same credit pool
- Codex integration: Coding agent tasks meter separately based on the rate card
- Workspace size: Larger teams naturally consume more credits through aggregate usage
- Monitoring tools: Admin dashboards help track consumption before credits run out
The July 6 transition forces financial discipline. Free agents encouraged experimentation. Paid agents demand ROI analysis.
| Plan Type | Credit Allocation | Additional Credits |
|---|---|---|
| Plus | Included monthly | Not available for purchase |
| Pro | Higher monthly allocation | Not available for purchase |
| Business | Shared workspace pool | Available for purchase |
| Enterprise | Custom allocation | Available for purchase |
| Edu | Education-tier rates | Available for purchase |
| Health | Healthcare-tier rates | Available for purchase |
| Gov | Government-tier rates | Available for purchase |
Organizations should analyze the 26-day free window data to project monthly costs. Teams that logged heavy agent usage during the trial period will see corresponding credit consumption once paid pricing activates.
How Do Workspace Agents Compare to Standalone AI Tools?
Workspace agents differ from standalone AI tools by integrating directly into organizational infrastructure, inheriting enterprise security policies, and operating within shared workspace environments. Standalone tools function independently, while workspace agents connect to synced applications, respect administrative controls, and operate through both ChatGPT and Slack interfaces.
The comparison reveals distinct trade-offs:
- Data governance: Workspace agents operate under Enterprise data policies; standalone tools may not
- Access control: Administrators manage who creates and uses workspace agents centrally
- Integration depth: Workspace agents connect to synced enterprise apps with in-region data support
- Collaboration: Agents can be shared across teams within the workspace
- Credit model: Workspace agents consume plan credits; standalone tools often use separate billing
- Slack presence: Workspace agents extend into Slack; standalone tools typically do not
- EKM compatibility: Workspace agents exclude EKM workspaces; standalone tools vary
- Audit capabilities: Enterprise workspaces provide logging; standalone tools may lack audit trails
Why would organizations choose workspace agents over standalone alternatives? The answer centers on control. Enterprise administrators gain visibility, security, and governance that standalone tools cannot match.
However, the EKM limitation creates a gap. Organizations requiring customer-managed encryption must look elsewhere or wait for OpenAI to add EKM support. This restriction affects financial institutions, government agencies, and healthcare providers that mandate encryption key control.
Frequently Asked Questions
Will workspace agents still function after July 6 without additional credits?
No. According to OpenAI’s release notes, ChatGPT workspace agents turn off when workspaces do not have additional credits after July 6. The agents cease functioning entirely rather than degrading partially, meaning organizations must maintain sufficient credit balances to keep agents operational.
Are ChatGPT workspace agents available for all Enterprise configurations?
No. The release notes explicitly state that workspace agents are not available for ChatGPT Enterprise workspaces with EKM (Enterprise Key Management) at launch. Organizations using customer-managed encryption keys cannot deploy workspace agents until OpenAI adds EKM compatibility.
How is Codex usage metered across different ChatGPT plans?
Codex credit rates vary across Plus, Pro, Business, Enterprise, Edu, Health, and Gov plans according to OpenAI’s published rate card. Each plan tier receives different credit allocations, and Business, Enterprise, Edu, Health, and Gov plans can purchase additional credits when base allocations deplete.
Can administrators restrict which users create workspace agents?
Yes. OpenAI’s admin quickstart documentation confirms that administrators configure roles and permissions, including who can create, share, and manage workspace agents. The admin console provides centralized control over agent creation, with granular settings for individual users and groups.
Summary
- July 6 deadline: Workspace agents require credits after this date — plan budgets accordingly
- EKM exclusion: Enterprise workspaces with customer-managed encryption cannot use agents yet
- In-region sync: All synced apps now support geographic data residency for compliance
- Administrative control: Admins govern agent creation, sharing, and monitoring through the console
- Credit budgeting: Analyze free-window usage data to project ongoing costs across plan tiers
Organizations deploying ChatGPT workspace agents should audit their free-period usage immediately. The July 6 transition leaves no room for uncertainty. Review the release notes, consult the Codex rate card, and ensure your workspace has sufficient credits before the deadline.