PatchArc vs Entire
The two closest tools for recording what an AI coding agent did: Entire stores full transcripts on a Git branch; PatchArc seals a signed, deterministic review capsule. Where each fits and where both can run together.
updated 2026-08-22 · competitor pricing and claims checked 2026-08-22 · PatchArc facts as of 0.2.0
Entire and PatchArc are the two tools on this site that answer the same question: what actually happened while an agent (or a person) made this change. They answer it with different artifacts. Entire captures the agent session itself, transcript and all, and stores it in Git. PatchArc captures Git activity plus what you chose to record, computes a review, and seals it so it can be verified later by someone who was not there.
At a glance
| Entire | PatchArc (0.2.0) | |
|---|---|---|
| What is recorded | Transcripts, prompts, files touched, token usage, tool calls, per commit | Commits and touched files from Git; your notes, decisions, risks, checkpoints |
| Agents | Claude Code, Codex, Copilot CLI, Cursor, Droid, Gemini CLI, OpenCode, Pi | Any tool that commits; transcript adapters not built yet |
| Artifact | Checkpoints on an in-repo branch (entire/checkpoints/v1) | A .parc ZIP with a signed manifest, stored in .patcharc/capsules/ |
| Review | The transcript is the record; grep it | Deterministic six-section review computed from evidence |
| Integrity | Git history | Ed25519 signature over a canonical manifest; per-file SHA-256; offline verify |
| Redaction | "Detects and redacts secrets before anything is stored" (vendor claim) | Engine with 13 detectors ships; not applied at seal yet |
| Sharing | Cloud hosting with regional options mentioned; pricing not published | ArcLink with random 96-bit slug, 30-day expiry, revoke and rotate; cloud re-verifies before publishing |
| License / price | MIT CLI; brew install --cask entire; no public pricing | Apache-2.0 CLI and format; Community free; Pro $12, Team $24 early access |
| Touches your branch | Never creates commits on your active branch | Never writes to Git at all; reads plumbing on a poll |
Facts about Entire come from its public repository and site as of 2026-08-22; we have not used a paid tier because none is published.
The core difference: transcript versus record
Entire keeps everything. If you need to know exactly what the agent was told, what it answered, and which tool calls it made, the transcript is the evidence, and keeping it in Git means it travels with the repository and is greppable by anyone with clone access. The cost is size and sensitivity: transcripts are large and contain whatever the agent saw, which is why Entire emphasises secret redaction before storage.
PatchArc keeps the shape of the work, not the conversation. The observer records commits and the files they touched; you add decisions and risks in one line each; stop computes a review (goal, what changed, affected scopes, verification, phases, decisions and risks) and seals it. The capsule is small, structured, and signed. The cost is that the conversation is gone: if the answer is in a prompt, PatchArc does not have it.
Verification
This is where the artifacts differ most. A PatchArc capsule carries its own public key and signature; patcharc verify checks the signature and every file hash offline and exits non-zero on any change, and the cloud repeats the check before publishing a link. A reviewer can trust that the record they are reading is the record that was sealed. Entire's integrity is Git's integrity, which is strong for history but is not a signature over a review, and nothing on Entire's public pages describes offline verification of a checkpoint.
Sharing
Entire's site describes hosted sharing with regional options; details and pricing are not public. PatchArc shares through an ArcLink: the cloud re-verifies the upload, the slug is random and unlisted, links expire in 30 days by default, and you can revoke (410) or rotate. Be aware that in 0.2.0 the ArcLink page is minimal and share defaults to public.
What Entire does better
- Captures the conversation, not just the commits. If your review question is "why did the agent do that", Entire can answer it.
- Covers eight agents out of the box today; PatchArc's agent adapters are not built.
- Stores in Git, so no new format to learn and no account anywhere.
What PatchArc does better
- A signed, portable, offline-verifiable artifact with a structured review, readable by someone outside the repository.
- Explicit human context: decisions and risks are first-class, timestamped entries, not lines buried in a transcript.
- Monorepo scopes: ten workspace detectors group the change by package.
- Server-side re-verification before anything is published.
When to choose which
Choose Entire when the transcript is the point: agent debugging, prompt audits, or a team that wants every session searchable in the repository.
Choose PatchArc when the artifact must be handed to someone who was not there and who needs to trust it: a reviewer, an auditor, a future you. Also when the repository is a monorepo and scope-level review matters.
Run both when you want the transcript for the people inside the team and the sealed review for everyone else. They do not conflict: Entire writes to its own branch, PatchArc writes only under .patcharc/. Record a patcharc note with the checkpoint id so the two artifacts reference each other.
Frequently asked questions
Does Entire sign anything?
Nothing on Entire's public repository or site describes signing or offline verification of checkpoints as of 2026-08-22. Integrity comes from Git history.
Can a reviewer verify a PatchArc capsule without an account?
Yes. patcharc verify <file> needs only the file and the CLI; the public key is inside the capsule. No account, no network.
What does each tool store, and where?
Entire stores checkpoints (transcripts, prompts, files touched, token usage, tool calls) on the entire/checkpoints/v1 branch in your repository, and optionally in its cloud. PatchArc stores a SQLite session database and sealed capsules under .patcharc/ in your repository, and uploads a capsule only when you run patcharc share.
Does PatchArc capture Claude Code transcripts?
Not in 0.2.0. It records the commits the agent makes and whatever you note. Transcript adapters are planned; see Working with AI coding agents.
Can I run both in the same repository?
Yes. Neither touches the other's files, and neither modifies your working branch.