ClineFlow
What is ClineFlow?
ClineFlow is an MIT-licensed workflow layer that keeps an AI coding agent's project knowledge (specifications, goals, decisions, verification evidence and task journals) as Markdown and YAML files inside your own repository, structured as a Google Cloud Open Knowledge Format (OKF) v0.2 bundle. It installs the same marker-delimited instruction block into the native project-instruction file of Cline, Claude Code, ChatGPT Codex, Cursor, GitHub Copilot and Windsurf, so context moves between agents, chats and teammates through ordinary Git workflows. The documented workflow runs no server, database or embedding service.
Last verifiedHow we evaluate
At a glance
- Developers and small teams running more than one AI coding agent against the same repository.
- Free. MIT-licensed and self-hosted, with no paid tier or hosted service.
What it actually installs
ClineFlow is a shell-script installer, not a hosted service. Run one prompt or a curl/PowerShell one-liner from a project directory and it drops a hidden .clineflow/ runtime (installer, updater, validators) plus a knowledge/ directory into the repo, then appends one marker-delimited instruction block to each supported agent's native config file: .clinerules for Cline, AGENTS.md for Codex and Cursor, CLAUDE.md for Claude Code, .github/copilot-instructions.md for Copilot, and .windsurf/rules/clineflow.md for Windsurf. Existing content in those files outside the marker block is left alone. There is no daemon, no account, and no network call at runtime beyond the install/update fetch from GitHub.
The memory model: five ledgers, not a vector store
Project memory lives as plain Markdown and YAML in the repo, not in an embedding index. Five small YAML files (clineflow_goals.yml, clineflow_specification.yml, clineflow_verification.yml, clineflow_last_session.yml, clineflow_timeline.yml) are meant to answer 'why, what, proven how, what's true now, and how did we get here' at a glance, and each substantial task gets a longer-form Engineering Journal file under knowledge/journals/. Saying 'please commit' triggers the agent to update the active journal, refresh all five ledgers with one shared timestamp, and run a bundled validator (validate-knowledge-sync --staged) before the commit is created. The format follows Google Cloud's Open Knowledge Format (OKF) v0.2 spec, a real, actively maintained Apache-2.0 project (9,100+ stars) for making agent-generated knowledge self-describing.
OKF conformance is self-declared, not certified
The README carries an 'OKF v0.2' badge and the repo ships its own OKF and knowledge-sync validators, but there is no independent body certifying conformance — the badge and the checks are both authored by the same maintainer. This is normal for how OKF is designed to work (no central registry, no required tooling) but it means 'OKF-compatible' here is a claim you can verify yourself by reading the spec against the repo, not one a third party has signed off on.
Maturity: near-solo, very early, and its own CI is red
The repo was created 2025-11-08 and has 1 star, 0 forks, and 0 open issues as of 2026-09-07. GitHub's contributor graph lists two accounts, but the second (aeonfluxor, 73 of the commits) is an unidentifiable account with no name, bio, company, or followers and only one public repo — there's no way to confirm it's a distinct person rather than the same maintainer's second account or an automated identity. Treat this as effectively a one-person project until shown otherwise. More concretely checkable: the last three runs of the project's own GitHub Actions test suite on main (as of 2026-09-07) all report conclusion: failure, driven by a windows-lifecycle-smoke job failing at its 'Install through Git Bash' step on every run in that window, even though Windows is one of the three officially supported install paths.
The marketing domain no longer hosts the product page
clineflow.com and www.clineflow.com both return HTTP 301 to https://hassanvfx.github.io/infinite-ai-context/, a GitHub Pages site that still displays the ClineFlow name and installs the same tool. Public RDAP records show clineflow.com was registered on 2026-08-20 — the redirect to GitHub Pages was therefore already in place well within a month of the domain's own registration. The GitHub Pages destination is legitimate (same author, same GitHub repo link, same install prompt) and not a broken link, but a buyer bookmarking clineflow.com should expect to land on a github.io URL, not the branded domain.
Cost
The tool itself is free: MIT-licensed, no account, no hosted tier, no usage limits, confirmed against the repository's LICENSE file and GitHub API metadata. The only paid artifact in the ecosystem is the author's companion book, 'Infinite AI Context: ClineFlow and Google's Open Knowledge Format' — $43.04 for a 159-page paperback on Lulu as of 2026-09-08, with a free web PDF also linked from the product page. Buying the book is optional and has no bearing on what the software does.
Frequently asked questions
What is ClineFlow?
ClineFlow is an MIT-licensed workflow layer that stores an AI coding agent's project knowledge (specifications, goals, decisions, verification and task journals) as Markdown and YAML inside your own repository, as a Google Cloud Open Knowledge Format (OKF) v0.2 bundle.
Is ClineFlow free?
Yes. It is MIT-licensed and runs from your own repository, with no paid tier and no hosted service. The author separately sells a companion book, Infinite AI Context, which is also published as a free web PDF.
Which coding agents does it work with?
It installs its workflow into the native project-instruction files used by Cline, ChatGPT Codex, Claude Code, Cursor, GitHub Copilot and Windsurf. The project states that no plugin, proprietary service or always-running process is required.
Does it need a database or an embedding service?
No. The documented workflow keeps knowledge as Markdown, YAML and links in Git, with no proprietary database or embedding service in the default setup.
What is the Open Knowledge Format?
OKF is an open specification published by Google Cloud that represents knowledge as Markdown files with YAML frontmatter. ClineFlow's knowledge bundle targets OKF v0.2 and ships validators that check the structure against it.
What happens if I remove it?
Removal runs as a dry run first, then transactionally deletes the managed runtime directory and the instruction blocks it owns. The knowledge bundle, generated dashboard reports and user documentation stay in the project.
How mature is the project?
The repository was created in November 2025, is MIT-licensed, publishes a Zenodo DOI, and was last pushed on 7 September 2026. It is actively developed and maintained by its author, Hassan Uriostegui.
Editor's read
Two things stand out once you look past the copy. The design is unusually easy to leave: the durable artefacts are Markdown, YAML and Git history in your own repository, and the uninstaller deliberately keeps the knowledge bundle when the tooling goes. The project is also young, with a repository dating from November 2025 and its author as sole maintainer, so the risk to weigh is continuity of the tooling rather than access to your own context. What would change our read is the commit loop breaking on a real project, or the bundle drifting away from the published OKF spec.
Explore other Agent Tools & Integrations
