Skip to content
Back to search

ContextLattice

The local-first intelligence layer that gives AI agents durable continuity, explainable retrieval, portable context, and verified learning across harnesses.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today

Check it deeper on your machine

It has changed since its scan on 11 Nov 2026.

As a Claude Code plugin, install it and then ask Claude Code:

/plugin marketplace add keithmackay/mackayi/plugin install scanskillsafety@mackayiDo a deep scan of https://github.com/sheawinkler/ContextLattice with scanskillsafety

Or from a terminal, clone it and run it:

git clone https://github.com/keithmackay/scanskillsafety-skill ~/.claude/skills/scanskillsafetynode ~/.claude/skills/scanskillsafety/cli.cjs --deep 'https://github.com/sheawinkler/ContextLattice'

It checks the exact commit: every file, its packages against OSV (api.osv.dev), how it compares with this page's scan (findsafeskills.com receives the repository's address), and, if you agree, Claude reading every file with your own Claude subscription. It never runs the code.

Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/sheawinkler--contextlattice/badge.svg)](/listings/sheawinkler--contextlattice)

HTML

<a href="/listings/sheawinkler--contextlattice"><img src="/api/listings/sheawinkler--contextlattice/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

Requirements: macOS, Linux, or Windows through WSL2; a Compose v2-compatible container runtime; and `gmake`, `jq`, `rg`, `python3`, and `curl`. The tested macOS baseline uses OrbStack through its explicit Docker context; see the [container runtime decision](docs/runtime/container-runtime-decision.md).

1. Install

git clone https://github.com/sheawinkler/ContextLattice.git
cd ContextLattice
cp .env.example .env
gmake quickstart

`gmake quickstart` is the prescribed technical install path; installers are bootstrap alternatives. The command prepares environment wiring, asks for a runtime profile, launches the selected local stack, and validates initial readiness.

2. Verify the runtime and retrieval path

curl -fsS http://127.0.0.1:8075/health | jq
contextlattice doctor --pretty
contextlattice state status --pretty
contextlattice context "verify this ContextLattice installation" \
  --project contextlattice \
  --pretty

Healthy containers are only the first check. The state command verifies the canonical gateway-owned storage inventory; the final command exercises the actual context path and reports source coverage, degradation, evidence, and next actions. Existing installs can use the explicit, reversible procedure in [gateway state migration](docs/runtime/gateway-state-root.md).

For a fuller lifecycle proof from the repository:

scripts/agent/agent-runtime-proof-pack --pretty
scripts/agent/agent-adoption-proof-matrix \
  --skip-provider-smoke \
  --progress \
  --pretty

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 324 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. # ContextLattice
  2. <p align="center">
  3. <a href="https://contextlattice.io/" target="_blank" rel="noopener noreferrer">
  4. <img src="docs/readme/contextlattice-editorial-hero.png" alt="ContextLattice editorial website hero showing a live local context field" width="100%" />
  5. </a>
  6. </p>
  7. <p align="center">
  8. <strong>Open an agent. Already there.</strong>
  9. </p>
  10. <p align="center">
  11. The local-first intelligence layer that gives AI agents durable continuity, explainable retrieval, portable context, and verified learning across harnesses.
  12. </p>
  13. <p align="center">
  14. <a href="#quickstart"><img src="https://img.shields.io/badge/Interface-CLI%20First-111111?style=for-the-badge" alt="CLI first"></a>
  15. <a href="https://github.com/sheawinkler/ContextLattice/releases/tag/v5.2.0"><img src="https://img.shields.io/badge/Release-v5.2.0-292929?style=for-the-badge" alt="ContextLattice v5.2.0"></a>
  16. <a href="#quickstart"><img src="https://img.shields.io/badge/Runtime-Local%20First-404040?style=for-the-badge" alt="Local-first runtime"></a>
  17. <a href="LICENSE"><img src="https://img.shields.io/badge/License-Apache%202.0-575757?style=for-the-badge" alt="Apache License 2.0"></a>
  18. </p>
  19. <p align="center">
  20. <a href="#quickstart">Quickstart</a> ·
  21. <a href="#how-it-works">How it works</a> ·
  22. <a href="#connect-your-agents">Agent setup</a> ·
  23. <a href="#architecture">Architecture</a> ·
  24. <a href="https://contextlattice.io/docs/">Docs</a> ·
  25. <a href="https://contextlattice.io/updates.html">Updates</a>
  26. </p>
  27. ## Stop replaying the brief
  28. Models can reason. Harnesses can act. Neither reliably retains the mission when a chat, model, tool, account, or computer changes.
  29. ContextLattice gives that work a durable, inspectable context layer. It reconstructs the active objective, selects the evidence that matters, carries it safely, and records what actually worked—without turning every prompt into a transcript dump or making cloud storage mandatory.
  30. | Capability | What changes |
  31. | --- | --- |
  32. | **Durable continuity** | Reopen the objective, decisions, repository state, risks, proof, and next move as one bounded packet. |
  33. | **Explainable retrieval** | Rank evidence by impact per token and expose source coverage, omissions, opposition, degradation, and receipts. |
  34. | **Portable context** | Move signed, least-privilege continuation across agents and machines while keeping execution and transport caller-owned. |
  35. | **Verified skill evolution** | Discover skills without loading every file, evaluate repeated wins on holdouts, and require review before promotion. |
  36. | **Privacy-bounded Aggregate Signal** | Learn from explicitly opted-in, clipped statistics while raw memory remains local; production activation stays hard-blocked pending independent privacy and utility review. |
  37. The **CLI is the primary interface**. The dashboard makes behavior and proof visible. HTTP and MCP are companion integration surfaces for applications and harnesses.
  38. ## How it works
  39. | Stage | ContextLattice does |
  40. | --- | --- |
  41. | **01 · Reopen** | Reconstructs the one active mission from durable checkpoints and current state. |
  42. | **02 · Select** | Retrieves high-signal evidence into a compact Context Pack with provenance. |
  43. | **03 · Move** | Carries signed, bounded context through Agent Packets, Passports, and encrypted continuation envelopes. |
  44. | **04 · Earn** | Records outcomes and promotes reusable behavior only after deterministic proof and human approval. |
  45. | **05 · Compound** | Improves future retrieval while preserving corrections, contradictions, freshness, and retirement semantics. |
  46. ContextLattice does not replace your agent harness, choose goals from retrieved text, or execute imported context. Local tools remain execution surfaces; memory and remote content remain evidence.
  47. [**Follow the evidence.**](docs/evals/evidence-chain-2026-11.md) Inspect the proof chain, dated results, public reproduction paths, and the claims those results do not establish.
  48. ## Quickstart
  49. Requirements: macOS, Linux, or Windows through WSL2; a Compose v2-compatible container runtime; and `gmake`, `jq`, `rg`, `python3`, and `curl`. The tested macOS baseline uses OrbStack through its explicit Docker context; see the [container runtime decision](docs/runtime/container-runtime-decision.md).
  50. ### 1. Install
  51. ```zsh
  52. git clone https://github.com/sheawinkler/ContextLattice.git
  53. cd ContextLattice
  54. cp .env.example .env
  55. gmake quickstart
  56. ```
  57. `gmake quickstart` is the prescribed technical install path; installers are bootstrap alternatives. The command prepares environment wiring, asks for a runtime profile, launches the selected local stack, and validates initial readiness.
  58. ### 2. Verify the runtime and retrieval path
  59. ```zsh
  60. curl -fsS http://127.0.0.1:8075/health | jq
  61. contextlattice doctor --pretty
  62. contextlattice state status --pretty
  63. contextlattice context "verify this ContextLattice installation" \
  64. --project contextlattice \
  65. --pretty
  66. ```
  67. Healthy containers are only the first check. The state command verifies the canonical gateway-owned storage inventory; the final command exercises the actual context path and reports source coverage, degradation, evidence, and next actions. Existing installs can use the explicit, reversible procedure in [gateway state migration](docs/runtime/gateway-state-root.md).
  68. For a fuller lifecycle proof from the repository:
  69. ```zsh
  70. scripts/agent/agent-runtime-proof-pack --pretty
  71. scripts/agent/agent-adoption-proof-matrix \
  72. --skip-provider-smoke \
  73. --progress \
  74. --pretty
  75. ```
  76. ### 3. Update Without Reinstalling
  77. For a managed installer release, check first and then apply through the same
  78. verified installer path:
  79. ```zsh
  80. contextlattice update --check --install-dir /path/to/managed-install
  81. contextlattice update --apply --install-dir /path/to/managed-install
  82. ```
  83. Normal apply checks running-runtime readiness and refreshes agent tools. Source
  84. checkouts keep their explicit Git/quickstart upgrade procedure; the updater never
  85. overwrites a checkout or crosses from paid to public. Windows applies its native
  86. MSI. See the [CLI update guide](docs/cli-updates.md) for scope and recovery.
  87. ## Connect your agents
  88. Run integration from the repository that should use ContextLattice:
  89. ```zsh
  90. cd /path/to/your/project
  91. contextlattice_adopt integrate \
  92. --repo . \
  93. --agents codex,claude-code,opencode,hermes-agent,hermes-ultra,omp,mercury-agent,pi,droid \
  94. --pretty
  95. contextlattice_adopt integrate \
  96. --repo . \
  97. --agents codex,claude-code,opencode,hermes-agent,hermes-ultra,omp,mercury-agent,pi,droid \
  98. --check \
  99. --pretty
  100. ```
  101. The integration command writes bounded managed blocks while preserving existing instruction text. It does **not** install Codex, Claude Code, OpenCode, Hermes, OMP, Mercury, Pi, Droid, or another third-party agent harness.
  102. External provider discovery is network-free. Provider execution remains explicit and caller-authorized; see the [external-provider boundary](docs/runtime/external-provider-auth-boundary.md).
  103. If an agent is performing the installation, it should follow the quickstart directly, avoid cloning a second checkout when already inside one, report the exact failing command and path, and rerun the deterministic check after any repair.
  104. ## The daily agent loop
  105. ```zsh
  106. # Confirm readiness when the environment is uncertain.
  107. contextlattice doctor --pretty
  108. # Retrieve scoped context before substantial work.
  109. contextlattice context "debug the current release regression" \
  110. --project contextlattice \
  111. --pretty
  112. # Save concise, durable progress.
  113. contextlattice remember \
  114. "Root cause verified; regression test added; focused checks pass." \
  115. --project contextlattice \
  116. --pretty
  117. # Resume without replaying the transcript.
  118. contextlattice resume --project contextlattice --pretty
  119. # Repair stale or wrong recall without silently rewriting history.
  120. contextlattice correct \
  121. "The prior deployment record is stale." \
  122. --category stale \
  123. --project contextlattice \
  124. --pretty
  125. # Close the loop with the verified outcome.
  126. contextlattice finish \
  127. "Regression fixed and verified." \
  128. --success \
  129. --project contextlattice \
  130. --pretty
  131. # Project the next bounded move or bind a completed response to durable proof.
  132. contextlattice_continuous_cognition status "prepare the next verified move" \
  133. --project contextlattice --session-id <session-id> --agent-id codex_gpt5 \
  134. --task-id <task-id> --objective-id <objective-id> --as-of <rfc3339> --pretty
  135. contextlattice_continuous_cognition evaluate "verify the completed response" \
  136. --project contextlattice --session-id <session-id> --agent-id codex_gpt5 \
  137. --task-id <task-id> --task-identity-id <task-identity-id> --as-of <rfc3339> --pretty
  138. # Prepare context for an external worker without exposing its one-shot claim.
  139. contextlattice agent-fit context-prep-schedule --project contextlattice \
  140. --session-id <session-id> --agent-id codex_gpt5 --payload-file prep-request.json --raw
  141. contextlattice agent-fit context-prep-claim --project contextlattice \
  142. --session-id <session-id> --agent-id codex_gpt5 --prep-id <prep-id> \
  143. --worker-id <worker-id> --claim-token-file prep.claim --raw
  144. contextlattice agent-fit context-prep-complete --project contextlattice \
  145. --session-id <session-id> --agent-id codex_gpt5 --prep-id <prep-id> \
  146. --claim-token-file prep.claim --payload-file prep-artifact.json --raw
  147. contextlattice agent-fit context-prep-use --project contextlattice \
  148. --session-id <session-id> --agent-id codex_gpt5 --prep-id <prep-id> \
  149. --task-id <task-id> --effective-profile-digest <sha256-digest> \
  150. --source-generation <generation> --raw
  151. ```
  152. Continuous Cognition is advisory-only: each invocation makes one bounded
  153. request, returns opaque evidence references, and never dispatches a runner or
  154. performs an external mutation. Context-preparation claims stay in an owner-only
  155. file and cross the completion/failure boundary only through the protected
  156. header; successful explicit use consumes the artifact once.
  157. Find a capability without loading every skill body:
  158. ```zsh
  159. contextlattice_skills_index search "browser automation" --pretty
  160. ```
  161. The active Skills Index scans configured Codex, Hermes, Hermes Ultra, and shared
  162. agent roots read-only. It reports each harness and root inventory separately,
  163. collapses byte-identical `SKILL.md` files by SHA-256 digest while retaining every
  164. source path as provenance, and requires discriminating query-term coverage
  165. instead of ranking generic words such as `skill`, `index`, or `agent`.
  166. Quarantine discovery remains separate, read-only by default, and never
  167. auto-promotes retrieved content.
  168. ## Architecture
  169. <table>
  170. <tr>
  171. <td width="50%">
  172. <a href="https://contextlattice.io/architecture.html">
  173. <img src="docs/public_overview/assets/architecture-service-map.svg" alt="ContextLattice service map" width="100%" />
  174. </a>
  175. </td>
  176. <td width="50%">
  177. <a href="https://contextlattice.io/architecture.html">
  178. <img src="docs/public_overview/assets/architecture-retrieval-flow.svg" alt="ContextLattice retrieval and learning flow" width="100%" />
  179. </a>
  180. </td>
  181. </tr>
  182. </table>
  183. The default local control path is:
  184. ```text
  185. Agent or application
  186. │
  187. ▼
  188. ContextLattice CLI / HTTP / MCP
  189. │
  190. ▼
  191. Gateway :8075
  192. ├── durable write and outbox fanout
  193. ├── scoped retrieval and source receipts
  194. ├── session, objective, graph, and outcome state
  195. └── dashboard-visible proof and operations
  196. ```
  197. Writes are validated and durably persisted before fanout. Retrieval merges the available sources, ranks bounded evidence, and reports missing or degraded coverage instead of hiding it.
  198. The active application path is Go and Rust. Python remains in build, development, migration, and audit tooling rather than the live request path. The exact runtime and toolset choices are recorded in the [v4 runtime decision](docs/runtime/v4-runtime-toolset-decision.md) and [container decision](docs/runtime/container-runtime-decision.md).
  199. ## Public and paid boundaries
  200. The public local lane is account-free and useful on its own. It includes the CLI-first memory lifecycle, Context Packs, sessions, graph and claim surfaces, Skills Index discovery, Agent Packets, public Passport and Mesh contracts, and local proof tooling.
  201. Paid artifacts add governed collaboration, protected activation, workspace operations, advanced analytics, and hosted distribution. They do not turn local memory into a mandatory cloud dependency.
  202. See [plans and distribution boundaries](docs/public_overview/premium.html) for the current contract.
  203. ## Install options
  204. macOS technical preview: unsigned DMG bootstrap launcher; expect Gatekeeper warnings until Developer ID notarization is configured, and prefer the source/CLI path.
  205. | Path | Best for | Status |
  206. | --- | --- | --- |
  207. | Source + `gmake quickstart` | Technical users and terminal-capable agents | **Recommended** |
  208. | `brew tap sheawinkler/contextlattice && brew install --cask contextlattice` | macOS convenience bootstrap | Available |
  209. | [macOS universal DMG](https://github.com/sheawinkler/ContextLattice/releases/latest/download/ContextLattice-macOS-universal.dmg) | Guided macOS bootstrap | Unsigned technical preview; expect Gatekeeper warnings |
  210. | [Windows x64 MSI](https://github.com/sheawinkler/ContextLattice/releases/latest/download/ContextLattice-windows-x64.msi) | Guided Windows bootstrap | Available |
  211. | [Linux bootstrap bundle](https://github.com/sheawinkler/ContextLattice/releases/latest/download/ContextLattice-linux-bootstrap.tar.gz) | Guided Linux bootstrap | Available |
  212. ### Resource profiles
  213. | Profile | CPU | RAM | Storage |
  214. | --- | --- | --- | --- |
  215. | Hugging Face / Glama lite | `2–4` vCPU | `4–8 GB` | `20–50 GB` SSD |
  216. | Local Lite core | `2–4` vCPU | `8–12 GB` | `25–80 GB` SSD |
  217. | Local Lite advanced | `4–6` vCPU | `12–16 GB` | `80–140 GB` SSD |
  218. | Local Full | `6–8` vCPU | `12–20 GB` | `100–180 GB` SSD |
  219. For heavier ingest, model storage, or the spike-lab adapters, read the [installation and storage guidance](https://contextlattice.io/installation.html) before selecting a profile.
  220. ## Security and privacy
  221. - Local-first and account-free in the public local lane.
  222. - API-key protection for operational routes.
  223. - Deterministic secret-like content filtering at write ingress: redact by default, block when configured, and allow only by explicit operator choice.
  224. - Provenance and trust isolation on retrieved memory.
  225. - Signed portable context and encrypted continuation envelopes.
  226. - Dry-run-first graph repair, source backfill, and quarantine workflows.
  227. - No automatic execution of retrieved instructions or imported continuation content.
  228. Security reports follow [SECURITY.md](SECURITY.md).
  229. ## Documentation
  230. | Need | Start here |
  231. | --- | --- |
  232. | Product overview | [contextlattice.io](https://contextlattice.io/) |
  233. | Installation | [Installation guide](https://contextlattice.io/installation.html) |
  234. | CLI and agent lifecycle | [CLI reference](https://contextlattice.io/cli.html) |
  235. | Harness and app integration | [Integration guide](https://contextlattice.io/integration.html) |
  236. | Architecture and scaling | [Architecture](https://contextlattice.io/architecture.html) · [Scaling memory](https://contextlattice.io/scaling-memory.html) |
  237. | Troubleshooting | [Troubleshooting guide](https://contextlattice.io/troubleshooting.html) |
  238. | Current behavior and release evidence | [Updates](https://contextlattice.io/updates.html) · [v5.2.0 release notes](docs/releases/v5.2.0.md) |
  239. | Roadmap | [Public roadmap](https://contextlattice.io/roadmap.html) |
  240. | Agent hooks | [Agent hook contract](docs/agent-hooks.md) |
  241. | Retrieval trust | [Retrieval receipts](docs/retrieval-receipts.md) |
  242. | Decision/session evidence | [Manual decision continuity](docs/decision-evidence-continuity.md) |
  243. | Skills and verified learning | [Skill efficacy review](docs/skill-efficacy-review.md) · [Skill Foundry](docs/outcome-policy-skill-foundry.md) |
  244. | Portable context | [Context Passport and Mesh](docs/context-passport-mesh.md) |
  245. | Local inference | [Local model options](docs/runtime/local-model-options.md) |
  246. | Full repository-backed manual | [Public field manual](docs/wiki/README.md) |
  247. The current public release baseline is [`v5.2.0`](https://github.com/sheawinkler/ContextLattice/releases/tag/v5.2.0).
  248. Commercial distribution follows its separately verified release; this public release does not authorize paid automation.
  249. ## License
  250. ContextLattice's public lane is licensed under the [Apache License 2.0](LICENSE).

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.