Entire: former GitHub CEO launches a Git forge for AI coding agents, with India hosting
Entire, led by former GitHub CEO Thomas Dohmke, has launched a Git hosting and code-review platform built for developers working with AI coding agents. Its 26 October announcement brings together native private repository hosting, a review system called Trails, and Continuous Memory for carrying context between coding sessions.
The useful part for Indian developers: India is one of its supported Git hosting regions. Non-Agent platform features are currently free, while Agent model usage is billed separately. There is a no-card trial credit, but it is not unlimited free AI.
What is Entire?
A Git forge is a service where teams store repositories, review changes and connect build and test systems. GitHub and GitLab are familiar examples. Entire offers another approach, focused on the large number of changes that coding agents can produce.
This is the launch of Entire's fuller forge platform, not the company's first appearance. Entire announced its company and its first open-source tool, Checkpoints, in February 2026.
At this launch, Entire supports native private repositories. Public native repositories and self-hosted Git nodes are still on its roadmap. Do not confuse mirroring an existing public GitHub repository with public native hosting on Entire.
Official announcement: https://entire.com/news/entire-launches-an-independent-git-forge-for-agent-human-collaboration
What changes with Trails?
A pull request normally brings together a code diff, comments and automated checks. Entire's Trails bring the task, code changes, agent sessions and review evidence into one place. A Trail can cover related changes across more than one repository.
On each push, configured review agents evaluate the changes. Entire describes evals for code quality, security, AI-generated clutter or "slop", and whether human review is required. Findings have severity levels and can include suggested fixes. The developer or coding agent can address them, then push another version.
Teams can enable Gates to require conditions before merging, including human approvals and passing CI checks. Entire integrates with Buildkite and Depot for builds and tests.
What this does not mean: a green score proves neither correctness nor security. Treat it as review evidence. Keep tests and human review appropriate to the project, especially for payments, authentication, hardware control and other high-impact code.
Official explanation: https://entire.com/blog/the-git-forge-for-agent-human-collaboration
What is Continuous Memory?
Coding sessions contain more than the final code: prompts, failed approaches, decisions and reasons for a change. Entire captures session context in Git so that later work can find it again.
Its two components have different jobs:
- entire-graph builds a map of the codebase, including definitions, callers and relationships, so an agent can find relevant code without repeatedly reading large parts of the project. Its built-in analysis runs locally without network requests or model calls.
- entire-brain links code context with past sessions and project memories. It can also connect to Entire's backend for shared team memory.
Entire reports up to 71% fewer tokens in its own testing, depending on the coding scenario. That is not a guaranteed saving on every task, and it does not mean that the whole hosted platform runs offline.
Source: https://github.com/entireio/entire-graph
Is it really 89 times faster than GitHub?
That headline needs context. Entire publishes a benchmark table showing 418 pushes per second, compared with 4.7 for GitHub and 65 for Cursor. These are Entire's own results using ForgeMark, its open-source benchmark.
ForgeMark measures sustained concurrent Git push throughput and latency under workloads such as many writers making small commits. This is not a promise that your laptop's next push, download or complete development workflow will be 89 times faster. Network distance, workload and concurrency matter.
There is another limit: if a repository is mirrored with GitHub as the source of truth, pushes are forwarded to GitHub and remain dependent on GitHub's speed and availability. Entire's earlier technical explanation says its high-throughput benchmark used Entire-native entire/unmirrored/ branches rather than the GitHub-backed write path.
AiCoN has not independently reproduced these benchmarks.
Sources: https://github.com/entireio/forgemark and https://entire.com/blog/an-entirely-new-git-hosting-network
What is free, and what costs money?
According to Entire's current billing documentation:
- Other platform features remain free; Agent model-token usage is billed through prepaid credits.
- Each billing account receives a one-time US$100 trial credit. It is service usage credit, not cash or a student cash benefit.
- No card is required for the trial. The credit expires after 90 days or when it is used up, whichever comes first.
- The trial starts when billing is set up or an Agent run starts, whichever happens first.
- Personal accounts and organizations have separate billing accounts. Credits cannot be transferred.
- Automatic top-up is off by default. At zero balance, Agents stop, but delayed metering and calls already running can cause a small overrun.
If you later buy credits, the documented minimum purchase is US$10 and purchased credits expire after 365 days. This is not a claim that pricing will stay unchanged forever. External services can add costs: Entire's launch blog describes Depot CI jobs as billed by the second.
The dollar values above are the service's credit units and purchase terms, not an INR cash offer. Check any rupee card charge and taxes before buying.
Current billing terms: https://docs.entire.com/billing
Why India hosting matters, and what it does not prove
Entire lists the US, EU, India and Australia as available Git hosting regions. Developers can choose a primary region and add regional mirrors.
A nearby region can be useful when developers or their agents operate there. But India Git storage does not, by itself, prove that every AI review or model request stays inside India. Do not treat it as a complete data-residency or compliance guarantee.
Repository guide: https://docs.entire.com/guides/repositories/entire-native-repositories
How to explore it without moving your main project
The following is a documentation-based starting path, not an AiCoN-tested walkthrough:
- Read the official quickstart and current billing terms first: https://docs.entire.com/quickstart.
- Use a small, disposable sample project. Remove real credentials, customer data, private prompts and sensitive screenshots.
- If using native hosting, sign in on the official site, open or create a project, choose its region, create a repository and copy the clone command shown for that repository.
- If choosing GitHub mirroring instead, review the Entire GitHub App's access carefully. Authorize only the repositories you intend to share. GitHub remains the source of truth for that mirror.
- Follow the quickstart for installing the CLI, enabling session capture for your supported agent, creating a Trail and configuring review Agents. Check what the hooks and settings will record before enabling them.
- Keep tests and human-approval Gates. Inspect the trial balance before running repeated Agent reviews. Do not add a card or enable top-up just to try the no-card credit.
For a college or team-owned repository, get the owner's permission before connecting it. A public tool launch is not permission to upload someone else's code or conversations.
Privacy: session memory can also preserve things you did not intend to share
Entire's documentation says it redacts recognized credentials before writing checkpoint metadata. However, PII redaction is disabled by default, and no redaction layer scans the contents of images. A screenshot containing a password or customer data may therefore reach the remote intact.
Review captured sessions and settings before syncing them. Local analysis in entire-graph should not be confused with cloud-based Trail review agents or shared session storage.
Privacy configuration: https://docs.entire.com/guides/configuration/privacy-and-redaction
Who is this useful for?
Entire is worth understanding if you already use coding agents and find that review queues or repeated loss of project context slow you down. Students can explore these ideas with a sample project and the no-card trial, while remembering that hosted Agent reviews consume credits.
If you only need basic Git hosting, there is no need to move an existing project because of a benchmark headline. At launch, native public repositories and self-hosting are still missing, and a move also means checking CI, access permissions and session privacy.
AiCoN's take: a real developer-platform launch with useful ideas, not a verified replacement for every GitHub workflow. The features and terms above were checked against official sources on 27 October 2026. AiCoN has not signed in, installed the software or tested the service for this explainer.
Official sources
- Forge launch: https://entire.com/news/entire-launches-an-independent-git-forge-for-agent-human-collaboration
- Detailed launch explanation: https://entire.com/blog/the-git-forge-for-agent-human-collaboration
- Billing and trial terms: https://docs.entire.com/billing
- Quickstart: https://docs.entire.com/quickstart
- Native repository hosting: https://docs.entire.com/guides/repositories/entire-native-repositories
- ForgeMark benchmark code: https://github.com/entireio/forgemark
- Earlier hosting architecture and benchmark scope: https://entire.com/blog/an-entirely-new-git-hosting-network
- Entire Graph: https://github.com/entireio/entire-graph
- Privacy and redaction: https://docs.entire.com/guides/configuration/privacy-and-redaction
- February company announcement: https://entire.com/news/former-github-ceo-thomas-dohmke-raises-60-million-seed-round