Anthropic
Terminal-first coding agent for repository editing, shell commands, code review, and long-running implementation workflows.
Reviewed
Selected releases, additions, and business disclosures, with links to the original sources.
Anthropic reports Claude wrote over 90% of its September code contributions. In Q3, the top 5% of Claude Code-using Enterprise users spent about 20 times the average for Enterprise users who did not use Claude Code. These are company-reported findings; the filing does not disclose updated Claude Code revenue or active users.
Version 2.1.287 lets plugins customize deeper agent behavior. Its optional You should know mod runs a side agent to flag missed issues; it requires a first-party session with telemetry enabled.
Attributed output
GitHub public-attribution count, 2026-06-09
12.5MAttributed outputGitHub public-attribution count, 2026-06-09Commits and PRs on public GitHub provably produced by this agent. Exact if it is visible through its own bot or workflow, an estimate if measured through attribution it declares (these can swing 30 to 40 percent between measurement runs at multi-million scale), and a floor either way.Index rank · all agents
#1 of 40 CLI agents
#1Index rank · all agents#1 of 40 CLI agentsRanked by agent-opened PRs in the trailing 4 complete weeks, matching the index. Agents without a tracked PR series follow, ordered by total attributed output. Other filters or sorts change the list position.Business signal
Claude Code run-rate revenue · Feb 2026
$2.5B+Business signalClaude Code run-rate revenue · Feb 2026$20/mo Pro
Entry price
Shared five-hour session and weekly usage limits
$20/mo ProEntry priceShared five-hour session and weekly usage limitsNot published
GitHub stars
No public repo
Not publishedGitHub starsNo public repo9,044,980
npm demand/wk
Adjusted from raw npm channel
9,044,980npm demand/wkAdjusted from raw npm channelWeekly package downloads as an install-demand floor; adjusted when the vendor's main install channel is no longer npm.576,387,762
npm installs to date
Since 2025-02-24
576,387,762npm installs to dateSince 2025-02-24Total npm downloads of this agent's package(s) summed across every tracked week since it was first published. A floor, like the weekly figure. Where a one-week automated/CI burst inflated the total, the baseline-only figure is shown too.From the public record
50 PRs · 50 repositories
Search window: – · captured Oct 14, 2026 (UTC)
Separate study: CodeRabbit follow-ups, May 25–October 11, 2026
In CodeRabbit’s May 25–October 11 audit, we found 51 original PRs crediting Claude, followed by 53 CodeRabbit auto-fix PRs; 39 follow-ups were merged. This cohort was selected through CodeRabbit, so it cannot tell us what fraction of this agent’s overall work receives follow-up fixes. Credits can name a model without identifying its client.
elizaOS/eliza: “fix(plugins): remove settings that no code reads” (merged). storytold/gridcraft: “io: remove the .sctmp temporary file when a save fails” (merged).
This is a sample of publicly attributed work, not total usage or a market-share estimate. Repositories are not customers. Attribution is declared in metadata, not independently verified tool execution; it does not prove exclusive authorship or autonomy. PR status is recorded at capture; merged does not establish correctness.
Up to 50 most recently created public PR search results within the audit window; post-filtered for body attribution or Codex branch. A recent convenience sample, not a random or exhaustive sample.
Returned records dated Oct 11, 2026–Oct 11, 2026.
Download the dated audit evidence (JSON)Weekly attributed output on public GitHub. Output visible through the agent's own bot or workflow is exact; output measured through declared attribution is an estimate. Window: 2025-02-03 to 2026-10-19.
Which Claude model its users actually run, from visible public attribution metadata on Claude-marked commits.
Weekly npm downloads since 2025-02-24, the install-demand trajectory.
Public GitHub commits attributed to Claude
Claude Code declares attribution on the work it produces, so coverage is strong. The headline is a point-in-time estimate that moves between measurements and a floor: squash-merges, private repositories, and a separate review integration sit outside it.
Last full week (ending 2026-10-25): 11.5M signed commits (+8.2% week over week), search estimate.
Distinct users: 8,875 humans opened the 12,600 most recently sampled PRs, about 1.4 PRs per person.
Merge rate: 91.6% of its pull requests get merged (8,764,896 of 9,573,369 over 8 settled weeks, 2026-08-24 to 2026-10-18). The window ends about ten days back so recently-opened PRs have had time to merge.
GitHub review activity: 3.6% of its merged PRs had at least one GitHub review record before merging; the other 96.4% had none. This is a broad weekly signal and can include bot reviews; the human-review read below comes from the enriched PR sample.
In significant repos (company-owned or 10+ stars), 30.5% of its merged PRs got a human review. Across all sampled repos, the rate is 20%. Human review means a non-author, non-bot GitHub reviewer in the PR-lifecycle sample. 1 sampled PRs had more than 100 review records, so those reviewer lists are marked truncated.
Median time to merge: 38 min. The unreviewed median is 0.36h. The human-reviewed median is 32.58h. From a 120-PR sample merged 2026-07-30 to 2026-08-28.
Test-inclusion: 47.5% of its sampled merged PRs touch a test file. This is a rough read on production readiness. 4 large PRs hit GitHub's 100-file path read cap, so treat this as a floor.
Flow share: 19.4% of the 1805 most recent public PRs sampled (350 PRs, 2 windows, last 2026-10-31).
Task lens: 4.2M agent-opened PRs. The declared mark can be removed by the team, so treat this as approximate.
Snapshot 2026-06-09, confidence Medium, tracked since 2026-06-09. Read the methodology.
Claude Code's npm downloads halved in late April 2026 — a distribution change, not a usage decline.
In late April 2026 Anthropic moved the recommended install to native installers (Homebrew, WinGet, apt, dnf, apk). The npm package still works and installs the same binary, so the drop from ~13M to ~8M weekly is installs shifting off npm, not fewer users. npm now counts only one channel and understates total adoption.
Trailing 4-week npm average. Native installs are unobserved, so this remains a floor.
Complete Mon-Sun week when cron-backed
Observed public npm range
As of 2026-10-25
| Signal | Value | Type | Confidence |
|---|---|---|---|
| npm downloads (2026-09-18 to 2026-09-24) | 11,936,572 | Public metric | High |
| Recommended install path | Native installer; npm remains supported | Official | High |
| Claude Code run-rate revenue | $2.5B+ as of February 2026; no updated product revenue in the October S-1 | Official | High |
| Claude Code business subscriptions | Quadrupled from Jan. 1 to February 2026 | Official | High |
| Claude Code enterprise mix | Enterprise use was over half of Claude Code revenue in February 2026 | Official | High |
| Internal code contributions written by Claude | Over 90% at Anthropic in September 2026; company-reported internal use | Official | High |
| Public GitHub commit share | SemiAnalysis estimated 4% of public GitHub commits were authored by Claude Code | Reported | Medium |
| Anthropic valuation | $965B post-money in May 2026 Series H | Official | High |
Share of Claude Code's own classified sampled work by segment; countries reflect where its PR authors are based. Full breakdowns: Languages, Frameworks, Industries, Map.
Pricing snapshot: 2026-10-22. Check the vendor for current terms.
Claude Pro is $20/month; Max is $100/month for 5x Pro's per-session allowance or $200/month for 20x, at web prices before tax.
Allowance is not a fixed token entitlement. Anthropic publishes no subscription token quota; per-session multipliers do not establish weekly token totals.
| Plan | Price | Allowance | Unit | Token / Overage Notes |
|---|---|---|---|---|
| Claude Pro | $20/mo | Claude Code included, with five-hour session and weekly limits shared with Claude. | session and weekly usage limits | Enable paid usage credits, purchase Console API credits, or wait for reset. Annual billing is $200 upfront. Pro includes Opus as well as Sonnet; no monthly API credits are included. |
| Claude Max 5x | $100/mo | Five times Pro's per-session usage allowance, resetting every five hours; weekly limits also apply. | per-session multiplier plus weekly limits | Enable paid usage credits, purchase Console API credits, or wait for reset. Monthly billing only. Claimable $100 monthly API credit requires a linked Console organization; new subscribers wait seven days. Credits expire each billing cycle and exclude interactive Claude Code and extra usage. |
| Claude Max 20x | $200/mo | Twenty times Pro's per-session usage allowance, resetting every five hours; weekly limits also apply. | per-session multiplier plus weekly limits | Enable paid usage credits, purchase Console API credits, or wait for reset. Monthly billing only. Claimable $200 monthly API credit requires a linked Console organization; new subscribers wait seven days. Credits expire each billing cycle and exclude interactive Claude Code and extra usage. |
Mean API-retail cost per completed task on the Artificial Analysis coding agent index (358 tasks across 3 suites, pass@1 over 3 runs, snapshot 2026-06-09). These are pay-per-token API costs, not subscription effective costs.
| Harness | AA index | Cost / task | Tokens / task | Time / task |
|---|---|---|---|---|
| Claude Code, Opus 4.8 (max) | 77.2 | $4.62 | 4.8M | 18m |
| Claude Code, Opus 4.8 (medium) | 67.2 | $1.60 | 1.7M | 9m |
| Claude Code, Opus 4.7 (max) | 66.6 | $4.14 | 5.7M | 14m |
| Claude Code, Opus 4.7 (medium) | 59.9 | $1.24 | 1.7M | 6m |
| Claude Code, Qwen3.7 Plus (thinking) | 53.1 | $4.98 | 3.3M | 11m |
| Claude Code, GLM-5.1 | 52.7 | $2.26 | 4.8M | 22m |
| Claude Code, Opus 4.6 (medium) | 51.3 | $1.27 | 2.2M | 7m |
| Claude Code, Kimi K2.6 | 50.5 | $0.76 | 3.7M | 42m |
| Claude Code, DeepSeek V4 Pro (high) | 50.1 | $0.35 | 3.5M | 18m |
| Claude Code, Sonnet 4.6 (medium) | 49.4 | $1.02 | 2.2M | 9m |
Break-even vs API retail on the top harness (Claude Code, Opus 4.8 (max), $4.62/task): Claude Pro at $20/mo covers about 4 tasks; Claude Max 5x at $100/mo covers about 22 tasks; Claude Max 20x at $200/mo covers about 43 tasks. Assumes the Artificial Analysis task mix.
Summary checked 2026-09-25.
Included with Claude Pro ($20/month), Max 5x ($100/month), and Max 20x ($200/month), at web monthly prices before tax. Claude Console API billing is separate.
Anthropic filed a public S-1 on October 26, 2026 and applied to list on Nasdaq as ANT. The preliminary filing leaves the offering price and share count blank; filing does not mean the IPO has completed. Claude Code is an Anthropic product.
Parent-company total. Claude Code is one of several Anthropic products, not a separately financed entity.
Anthropic does not train models on code sent under commercial terms. Enterprise customers can add a zero-data-retention addendum, and Claude Code can route through Amazon Bedrock or Google Vertex in a customer cloud account.