AXLR vs Claude Code

AXLR 60

No payment protocol (0). Free and Apache-2.0 with nothing to buy, so 20, 20 and 20 on the last three lines by the self-hosted rule, since the docs say what you pay your model provider, the free route needs no card, and it runs local models with no signup at all, the JSON worker without any model. 0+20+20+20 = 60.

Claude Code 20

Harness reading of the published rubric. No payment protocol (0). Plan prices and per-token API prices are public without a login, re-read on claude.com/pricing on 17 October, though plan usage limits aren't given as numbers (20). No free route. The free claude.ai plan excludes Claude Code, and the API, Bedrock, Google Cloud and Foundry are paid (0). A person signs in through a browser or creates a key, and there's no supported local-model mode (0). 0+20+0+0 = 20. Judgement call, against Anthropic's own product. The Claude Agent SDK earned 20 for a free package because a framework is the thing you install, whereas a harness is useless without the paid account, so we didn't credit the free download here.

AXLR 55

Apache-2.0 with a NOTICE (30). No privacy policy or terms on underpassai.com or axlr.underpassai.com; the docs' 'Where data goes' and 'Diagnostics and privacy' sections say that model requests go to OpenRouter, the local server or the native API, that Jev sends text to TypeSafe only when on, that sessions live under $XDG_STATE_HOME/axlr, that payload capture is off by default and redacted when on, and that traces are deleted after 30 days; nothing covers the engine downloads' or the catalog registry's handling of a client's requests (15). No deprecation policy; pre-1.0 changes are dated inline in the docs and the product contract records the 5 October 2026 retirement of the first skill catalogue after the fact (6). No telemetry to disclose, and the docs say in two places what leaves the machine and how to stop payload capture, though no single page says there is no telemetry (17). 30+15+6+17 = 68.

Claude Code 81

Closed source with clear terms, LICENSE.md pointing to the Commercial Terms, unchanged on 17 October (15). The data-usage page, re-read on 17 October, states retention per account type (consumer 30 days or 5 years with training on, commercial 30 days, zero retention for qualified Enterprise), 30 days of plaintext local transcripts, 5 years for /feedback reports and 6 months for shared survey transcripts, and these agree with the privacy policy links, but the logging and error-tracking services are named only as third parties. The page doesn't say that since 2.1.300 local transcripts also record each API request's system prompt, tools and parameters. We read that as an omission rather than a contradiction, since it describes the same local files (25). No written deprecation policy. The README still marks npm installs deprecated without a date, while the setup page lists npm without that label. The changelog page dates each release after it ships, and the one dated advance notice found, standalone Claude Design closing by 14 December 2026 in 2.1.302, concerns a separate product (8). Telemetry is documented per provider and per service with an opt-out for each, re-read on 17 October (20). 15+25+8+20 = 68.

AXLR 70

Read as a local package, as claude-code, goose, opencode and openhands were. Release archives for Linux, macOS and Windows on amd64 and arm64 with a .sha256 beside each, a GHCR image, an OCI Helm chart and source builds with Go 1.26, all stated in the docs; no Homebrew, npm or apt route (20). CI on main runs vet, race tests and a coverage gate above 80 per cent for both modules, with the sandbox tests required under bubblewrap, and the last ten runs on main to 1 November 2026 all passed; the release workflow tests on six native runners (25). 10 open issues, all opened by the founder on 19 and 21 October 2026 as trackers for service work (work modes, plugins, the sandbox setting, cost limits), none from an outside user and none a crash report, 0 open pull requests and about 100 merged; with 0 stars there is nobody else to report, so the handling of outside reports is untested (20). Tags are semver 0.x and the docs date each behaviour change inline ("until 14 October 2026…"), but the 21 releases have empty notes, there is no CHANGELOG.md, and the product contract records one retirement (the seven 1.0 skill definitions on 5 October 2026) without a notice period (5). Pre-1.0, which the README calls an evolving project (0). 20+25+20+5+0 = 70.

Claude Code 50

Local-package reading. Official installers for macOS 13+, Windows 10 1809+, Ubuntu 20.04+, Debian 10+ and Alpine 3.19+, signed apt, dnf and apk repositories and a GPG-signed manifest of checksums for every binary, re-read on the setup page on 17 October (20). No public CI or test suite for the program, since the source isn't published, and the repository's workflows on 17 October are still issue triage and tests for mods (0). Over 5,000 open issues and 735 open pull requests, with area, platform and regression labels on new reports but no maintainer replies visible on the first page, per the 1 October check, not re-read on 17 October (10). Every change still ships as a 2.1.x patch. 2.1.299 narrowed MCP allow rules so mcp__a no longer covers a server named a__b, 2.1.300 made two removed claude remote-control flags exit with an error, and 2.1.302 on 16 October changed the starting permission mode of claude -p and Agent SDK sessions again, with no breaking-change heading, though a stable channel skips releases with major regressions (5). 2.1.302 (15). 20+0+10+5+15 = 50. Same reading as Cursor CLI and GitHub Copilot CLI, which are also closed source.

AXLR 62

Harness reading of the framework checklist, as for the other four harnesses. 30 for what leaves the machine by default, 20 for approvals and sandboxing, 15 for prompt-injection posture, 15 for audit and 20 for the security programme. No product telemetry in the console source (the outbound hosts are openrouter.ai, api.anthropic.com, api.openai.com, a configured local server, api.github.com and github.com for engine downloads, and registry.underpassai.com for the signed catalog index) or in the docs; payload capture is off by default since 9 October 2026, traces omit prompts and credentials, model keys are never given to commands or plugins, and Jev sends requests and answers to TypeSafe only when on, which the setup assistant does when TYPESAFE_API_KEY is set, with a notice on stderr (27). Approval cards, a destructive list that always asks, four work modes, a policy level per execution, a bubblewrap or container sandbox with an enforced egress allow list, and a service that approves nothing but reads. Against that, the setup assistant writes autonomous: true, exec_sandbox is off by default, forged tools run with the person's rights, and the README says the console and worker are not sandboxes (13). Forged tools are refused when a file changes since the forge, a policy forbids calls outside the model, and the docs say MCP tools keep their own side effects, but no prompt-injection guidance was found and MCP results reach the model unmarked (6). The service keeps an event journal, audit records, a receipt with the SHA-256 of every call's arguments and outcome, and correlation ids on every event; the console keeps session records, a redacted app log and /session (14). No SECURITY.md, no security.txt on either host, no advisories, no bug bounty; release checksums and a minisign-signed catalog are integrity, not a programme (2). 27+13+6+14+2 = 62.

Claude Code 80

Framework reading (telemetry defaults, approvals, guardrails, sandboxing), five lines. Usage metrics on by default on the Claude API, without code, prompts or file paths, error reports on by default for Pro and Max sign-ins since v2.1.198, and every provider sends WebFetch hostnames to api.anthropic.com for a blocklist check, each with an opt-out. Metrics, error reports and /feedback are off by default on Bedrock, Google Cloud, Foundry and Claude Platform on AWS, while session surveys and the WebFetch check run on every provider. Consumer plans train on data when the user's setting allows it. The data-usage page was re-read on 17 October (18). Six modes, allow, ask and deny rules, protected paths, and managed settings that can disable bypass and auto mode and that nothing else overrides. The sandbox is opt-in and absent on native Windows, and auto mode, a classifier rather than a person, is the starting mode for interactive sessions, all re-read on the permission-modes and sandboxing pages on 17 October (16). The auto-mode classifier blocks named exfiltration patterns, the sandbox proxy denies hosts outside an allowlist that starts empty, and WebFetch checks a blocklist, but since 1 October 2026 a mod can approve a call a deny rule refuses unless managed settings or a Team or Enterprise sign-in hold it (11). Opt-in OpenTelemetry with tool content behind OTEL_LOG_TOOL_CONTENT, local transcripts for 30 days, and cloud sessions logged through a proxy. 2.1.301 adds classifier_verdict and permission_mode to tool_decision events (15). HackerOne bug bounty and SECURITY.md, both unchanged in the repository on 17 October, advisories published with GHSA ids, the newest GHSA-pq7j-f95f-qfcv (High) on 15 October, and anthropic.com's security.txt valid to 31 December 2026, re-read on 17 October (20). 18+16+11+15+20 = 80. SOC 2 isn't scored on the framework reading. Advisories, and the sandbox and permission-rule fixes in 2.1.288 to 2.1.300, count under negative events. The mods held behind deny rules are as described in the sec-default README re-read on 15 October. Judgement call, written by an agent on Claude, and this is the same checklist the other four harnesses got.