Verified AI Launch Profile
GitHub Copilot AI Agent Session Limits
GitHub added public-preview AI credit session limits to Copilot CLI and the GitHub Copilot SDK on July 1, 2026.
At a glance
Launch Snapshot
- Company
- GitHub
- Launch date
- July 1, 2026
- Launch type
- Major Update
- Category
- AI Coding Tools
- Pricing
- GitHub says session limits require Copilot CLI 1.0.66 or later or Copilot SDK 1.0.5 or later and are available in public preview for Individual, Business, and Enterprise plans. The limit is a soft cap because a response already in progress can finish after the threshold is crossed; current plan pricing and AI credit terms remain authoritative.
- Free plan
- Unknown
- API
- Yes
- Open weights/source
- No
Launch Context
Use these links to move from this record into the broader Launch Intelligence database.
Verification & Sources
- Status
- Verified
- Source links
- 3
- Freshness
- Verified July 9, 2026
- Last verified
- July 9, 2026
- Last updated
- July 9, 2026
Key source checks
Suggest a correction
Creator Coverage Next Steps
This launch has signals that may support demos, reviews, creator education, founder storytelling, or practical product explainers.
Launching an AI product that needs clear demos, creator education, and buyer trust? Sponsor a Kingy AI video or launch feature.
Kingy AI Take
Copilot AI credit session limits are a practical governance control for unattended agent runs because they put a task-level boundary around model, subagent, and background usage. The soft-cap behavior and public-preview status mean teams should treat the feature as one layer in a broader budget and monitoring policy rather than a hard billing guarantee.
Who it is for
Developers, automation teams, engineering leaders, and enterprise platform teams that need per-session controls for unattended or long-running Copilot agent work.
What feels promising
A session-level spending boundary gives interactive users and automated jobs a direct control that complements organization budgets without forcing a task restart after an interactive limit adjustment.
What feels unproven
The feature is in public preview, the cap can be exceeded by a response already in flight, and organizations must still combine it with broader budgets, policy, monitoring, and workload-specific tests.
Source-backed record