KODEATLASby Kalu Kode
Discuss early access

Current product · KodeAtlas CLI v0.2.0

A verifiable evidence archive for related GitHub repositories.

The current product is a production-oriented, read-only archival CLI that preserves Git source history and GitHub collaboration history together, with provenance, completeness, and integrity.

StatusImplemented current baseline
RuntimeStrict TypeScript · Node.js 22+ baseline
Provider behaviorRead-only Git and GitHub collection
kodeatlas initkodeatlas exportkodeatlas updatekodeatlas verifykodeatlas inspect

Evidence collected today

Source history and collaboration history stay connected.

The archive keeps raw source evidence alongside stable normalized records so future normalization, local browsing, transcript correlation, and static generation do not have to trust one lossy export.

Current

Git source history

Bare mirror, branches, tags, pull-request refs, commits, trees, blobs, object inventory, fsck output, verified Git bundle, and optional Git LFS objects.

Current

Pull requests

PR metadata, comments, reviews, review comments, review threads, checks, statuses, local/API reconciliation, patches, changed files, timelines, and attachments.

Current

Issues and planning metadata

Issues, comments, events, labels, milestones, reactions, cross-reference evidence, state reasons, and attachments.

Current

Releases

Release metadata, notes, tag/commit relationships, uploaded assets, hashes, and release-note attachments.

Current

Provenance

Raw REST and GraphQL responses, request metadata, pagination, API versions, response hashes, source pointers, and rate-limit headers.

Current

Indexes

Entity, commit, activity, relationship, and attachment NDJSON indexes.

Current

Integrity and completeness

SHA-256 manifests, Git-bundle verification, referential-integrity checks, explicit partial states, warnings, statistics, and immutable run receipts.

Efficiency model

Use Git for complete source history. Reserve APIs for collaboration evidence.

KodeAtlas uses the local Git object graph for commit history, pull-request commit ranges, changed-file inventories, and patches whenever the required objects are available. GitHub APIs are reserved for collaboration metadata and fallback reconciliation.

This avoids treating capped provider endpoints as complete history and reduces unnecessary API consumption.

Resumability

Long-running collection should fail honestly and resume safely.

  • conservative request scheduling;
  • repository-wide grouping instead of repeated per-item requests;
  • ETag and Last-Modified revalidation;
  • primary and secondary rate-limit handling;
  • atomic replacement and durable checkpoints;
  • signal-aware cancellation and resumable exit states;
  • structured progress with a heartbeat and request/cache/retry counters.

What the archive is not yet

The evidence foundation is current. The living context layer is next.

KodeAtlas CLI v0.2.0 does not yet track functions or classes through refactorings, import Codex or Cursor transcripts, connect backlog work to semantic anchors, infer regression relationships, coordinate agents, or serve live MCP context.

Those are explicitly later phases. The archive becomes their source-evidence layer rather than being discarded or reimplemented.

See what comes next

Take the next step

Start with evidence that will still be useful later.

The archive foundation can stand alone while later intelligence layers are evaluated progressively.

View capability status Discuss the context platform