Anthropic Claude Code vs Factory (2026)

Upgraded from P: terminal, desktop, IDE, web, GitHub Actions, Remote Control from mobile, and a headless -p flag for scripting are seven distinct invocation paths, plus hooks firing on lifecycle events.

Upgraded from P: invocation spans app, CLI, headless exec, cloud computers, mobile, Slack, and ticket systems as first class entry points, plus automations and CI. Ticket driven entry is the vendor's distinctive claim and it is a genuine trigger surface, not just an integration.

Upgraded from P: a Knowledge layer indexing repo, docs and ticket history, plus wikis exposed through the API and AGENTS.md configuration, is grounding on the customer's own material across several sources rather than repository search alone.

Upgraded from P: CLAUDE.md is durable repository scoped memory read every session, session transcripts persist and are resumable, and Remote Control syncs a transcript across devices.

Upgraded from P: cloud session sync across desktop, web and mobile plus a persistent Knowledge layer and wikis is durable cross session and cross surface state, not just context within a run.

Held at F on the documented enterprise surface, but confidence lowered to medium: the security and identity documentation was reached only through the docs navigation and the DeepWiki mirror rather than a first party security or trust page, and no attestation page was retrieved directly on either pass. SOC 2 Type II is reported consistently by third parties. Worth a dedicated fetch of a trust page next pass.

Upgraded from P: OpenTelemetry export with tool level detail is an organisation grade audit trail, and cloud sessions log all operations for compliance. Hooks make the trail programmable, which is stronger than the reporting-only surfaces I downgraded elsewhere in this lane.

Downgraded from F. Analytics and telemetry are documented as an enterprise capability and the API exposes analytics, but that is fleet level usage reporting rather than a per action execution trace reconstructing why a droid acted. Same reading applied across this lane; cursor held at F because it documents an AI code tracking API and audit log as distinct products, which was not retrievable here.

Upgraded from P: self hosted environments for cloud sessions and customer chosen inference platforms are both documented, alongside the default of running entirely on the developer's own machine. Recorded honestly: the docs state data is encrypted in transit but not at rest, and Zero Data Retention is an enterprise arrangement rather than a default.

On premises deployment is documented as an enterprise option alongside cloud Droid Computers and fully local CLI execution, which is genuine deployment choice. Contrast cursor, reviewed immediately before this, which explicitly offers no on premises path.

Upgraded from N: plugins are a versioned installable unit distributed through a public marketplace with a discovery and browse UI, plus bundled skills shipped by Anthropic. That is the axis, and it did not exist in this form when the June record was built.

Upgraded from N. The axis measures whether the customer chooses the model powering the product. Bedrock, Vertex and Foundry are customer selected inference platforms under the customer's own cloud account and credentials, and model selection plus fallback-model routing are documented, with hooks able to block or annotate a model switch. All models are Anthropic's, which is a real constraint and is recorded here, but provider and deployment plane choice plus per subagent model selection clears the F bar as applied to activepieces and coderabbit in this lane.

Model independence is a named platform capability with BYOK documented, so the customer both selects the model and can bring their own provider credentials; enterprise model policies and gateways let an organisation constrain that centrally.

Downgraded from F, applying the same axis rule used on coderabbit, kiro, google-antigravity and jetbrains-ai in this lane. /verify, /run and /code-review test the customer's code and confirm the agent's change works, which is self verification of output; no harness for evaluating or regression testing agent behaviour itself is documented. Graded consistently rather than favourably.

Downgraded from F per the lane wide axis rule. QA automation and automated code review are droids acting on the customer's code, and the Agent Readiness framework assesses the customer's codebase, not agent behaviour. No harness for evaluating or regression testing the agents themselves is documented. Consistent with coderabbit, kiro, jetbrains-ai, google-antigravity, Claude Code and cursor.

Meets the axis test squarely: the agent drives real web applications through their human interface, using the customer's existing logged in session, reaching software that exposes no programmatic interface to it. Confirmed on Anthropic's own docs rather than the unverifiable sub-agent browser surface reference the June basis cited.

Downgraded from F, same correction as cursor from this batch. The May basis had no URL and no content. Third party reporting says the Desktop app grants supervised access to a developer's browser, which would meet the axis, but no first party page documenting a browser tool was retrieved across two passes, and third party sources cannot establish a capability. Graded as not documented; this is the most likely cell to move back if a Desktop app docs page turns up.