ultraworkers/claw-code
59.4
Adequate · 14 November 2026
110.4k
lines of production code
Rust
primary language
6
measurements over time
What this system is
Features
Add default configuration files and project documentation
The repository now ships with default configuration files (\.claude.json\ and \.claw.json\) and a comprehensive knowledge base (\AGENTS.md\) to guide contributors and automated agents. The \.claude.json\ sets the default permission mode to \dontAsk\, while \.claw.json\ defines command aliases. \AGENTS.md\ provides a detailed map of the Rust workspace structure, code conventions, and known anti-patterns. Additionally, a \.dockerignore\ file has been added to optimize Docker build contexts by excluding unnecessary files and directories.
(repo-wide) · high confidence
Added Asolaria wire benchmark example
A new reproducible benchmark example (\examples/asolaria-wire\) has been added to compare JSON text serialization against a fixed-width binary format with SHA-256 tamper-evidence. This tool measures wire size, encode/decode speed, and integrity verification for agent-lane-event records, demonstrating that the binary approach offers smaller uncompressed wire size and faster decoding while providing per-record tamper detection that JSON lacks natively.
examples/asolaria-wire · high confidence
Added bundled plugin scaffolds for hook integration testing
Added two new bundled plugin scaffolds, \example-bundled\ and \sample-hooks\, to the Rust plugin system. These plugins are disabled by default and include \PreToolUse\ and \PostToolUse\ shell hooks, providing a concrete structure for testing hook integration.
rust/crates/plugins/bundled · high confidence
Added compatibility harness for extracting CLI manifest from upstream source
The \compat-harness\ crate has been introduced to parse and extract command, tool, and bootstrap configurations from the upstream TypeScript source files (\src/commands.ts\, \src/tools.ts\, \src/entrypoints/cli.tsx\). This allows the Rust runtime to automatically detect available commands (including internal-only and feature-gated ones), tools, and bootstrap phases without requiring manual synchronization, supporting various repository layouts via environment variables and path resolution.
rust/crates/compat-harness · high confidence
Added installation script and mock parity testing harness
The rust/scripts directory now includes an install.sh script that builds the release binary and links it to \~/.local/bin/claw, simplifying local installation. Additionally, a new mock parity testing infrastructure has been added, consisting of a Python harness (run\_mock\_parity\_diff.py) and a shell wrapper (run\_mock\_parity\_harness.sh) that execute the mock\_parity\_harness tests to verify behavioral consistency against defined scenarios.
rust/scripts · high confidence
Automated lane completion detection and PDF text extraction
The tools crate now includes a \lane\_completion\ module that automatically detects when a lane is finished by verifying the agent output status is 'completed' or 'finished', there are no errors or blockers, tests are green, and code has been pushed; this triggers policy actions to close out the lane and clean up the session. Additionally, a new \pdf\_extract\ module provides minimal text extraction from PDF files by parsing content streams and decompressing FlateDecode data, allowing the system to read text from PDF documents without external dependencies.
rust/crates/tools/src · high confidence
Build-time version provenance injection
The \claw\ binary now embeds detailed build metadata at compile time, including the Git SHA, short SHA, branch name, dirty status, commit date, timestamp, Rust compiler version, target triple, and build date. This is implemented via a new \build.rs\ script that queries \git\ and \rustc\ during the build process, enabling users to verify the exact provenance and reproducibility of their installed binary.
rust/crates/rusty-claude-cli · high confidence
CC2 board schema and rendering now executable for G001
The \.omx/cc2\ area now contains a canonical board (\board.json\ and \board.md\) that maps every frozen ROADMAP heading and ordered action to verifiable items. This board is structured to be executable by the G001 human board and docs rendering pipeline, providing a single source of truth for lifecycle status, release buckets, and stream ownership across the Claw Code 2.0 roadmap.
.omx/cc2 · high confidence
Expanded CLI for runtime simulation and session management
The \src\ porting workspace CLI now supports new subcommands that simulate runtime behaviors and manage session state. Users can inspect the command and tool graphs (\command-graph\, \tool-pool\, \bootstrap-graph\), run a simulated turn loop (\turn-loop\), and persist or load session transcripts (\flush-transcript\, \load-session\). Additionally, the \commands\ and \tools\ subcommands now accept flags to filter out plugin/skill commands or MCP tools, and new \exec-command\ and \exec-tool\ shims allow users to simulate command execution against the mirrored inventories.
src · high confidence
Introduce Rust CLI command registry and slash command surface
The \rust/crates/commands/src\ crate now provides the core command registry and slash command definitions for the CLI. This includes a \CommandRegistry\ for managing command sources (Builtin, InternalOnly, FeatureGated) and a comprehensive list of \SlashCommandSpec\ entries covering session management (help, status, compact, clear, resume), configuration inspection (config, mcp, memory, permissions, model), cost tracking, and git-integrated workflows (diff, commit, pr, issue, bughunter). This change establishes the structural foundation for the CLI's command dispatch and help system in Rust.
rust/crates/commands/src · high confidence
Introduce Rust-based CLI with interactive setup wizard and project initialization
The Rust CLI (\rusty-claude-cli\) is now available as a functional replacement for the legacy CLI, featuring a new \claw\ binary. It includes an interactive setup wizard (\setup\_wizard.rs\) to configure API providers (Anthropic, xAI, OpenAI, DashScope), keys, and models, and a project initialization command (\init.rs\) that scaffolds \.claw/settings.json\, \.claw.json\, and \CLAUDE.md\ while updating \.gitignore\. The interactive REPL uses \rustyline\ for line editing and command completion, and terminal output is rendered with \crossterm\ and \pulldown\_cmark\ for rich, colored Markdown and spinner animations.
rust/crates/rusty-claude-cli/src · high confidence
Introduce \`claw-analog\` minimal agent harness
Adds a new \claw-analog\ binary and library crate providing a lean tool-agent loop (read/list/grep/write) on claw-code API providers. The harness includes a \doctor\ command for environment and build sanity checks, a \config validate\ command to parse and preview \.claw-analog.toml\ merges without API calls, and an \agents\ command to run multiple specialized sub-agents (audit, explain, implement) sequentially with configurable permissions and presets. It supports streaming output, JSON events for CI, and non-interactive permission enforcement.
rust/crates/claw-analog · high confidence
Introduce mock Anthropic service for behavioral parity testing
A new \mock-anthropic-service\ crate has been added to provide a local TCP server that mimics the Anthropic API for testing purposes. This service supports multiple behavioral scenarios—including streaming text, file operations, bash execution, tool usage, auto-compaction, and token cost reporting—allowing the CLI to verify its interactions against a controlled mock. The service captures incoming requests and returns scenario-specific responses, enabling parity checks between the CLI and the real API without external dependencies.
rust/crates/mock-anthropic-service · high confidence
Introduce plugin subsystem with hooks, tools, and registry
The \rust/crates/plugins\ crate now provides the core infrastructure for the plugin system, including a plugin registry that discovers and validates plugins from builtin, bundled, and external sources. Users can now define and execute pre- and post-tool-use hooks via shell commands, with the \HookRunner\ managing event propagation and error handling. The system also supports plugin-defined tools with permission levels (read-only, workspace-write, danger-full-access) and lifecycle hooks (init/shutdown). A test isolation utility (\test\_isolation.rs\) ensures plugin tests run in isolated temporary environments to prevent state leakage.
rust/crates/plugins/src · high confidence
New Rust API client with configurable timeouts, proxy support, and prompt caching
The API layer now uses a new Rust client (\rust/crates/api/src\) that introduces configurable HTTP timeouts via \CLAW\_API\_CONNECT\_TIMEOUT\ and \CLAW\_API\_REQUEST\_TIMEOUT\ environment variables, standard proxy support (\HTTP\_PROXY\, \HTTPS\_PROXY\, \NO\_PROXY\), and a local prompt cache for Anthropic models to reduce latency and cost. The client also supports multiple providers (Anthropic, OpenAI, XAI/Grok) with automatic routing, enriched error reporting for JSON parsing and authentication, and structured thinking block handling.
rust/crates/api/src · high confidence
New provider abstraction layer with native support for Anthropic, OpenAI-compatible APIs, and Ollama
The API layer now uses a unified \Provider\ trait to route requests to specific backends, introducing native support for Anthropic (via \ANTHROPIC\_API\_KEY\), OpenAI-compatible endpoints (including xAI, Alibaba DashScope, and local Ollama instances via \OLLAMA\_HOST\), and a registry of model aliases (e.g., \opus\, \sonnet\, \grok\). This change adds configurable retry policies, request body size pre-flight checks to prevent hard failures, and specific request sanitization for each provider (such as stripping Anthropic beta fields or converting stop parameters).
rust/crates/api/src/providers · high confidence
New runtime subsystems for approval tokens, bash execution, and configuration validation
The runtime crate introduces several new capabilities: a structured approval-token system for policy exceptions and delegation; a bash execution tool with sandboxing, timeout, and provenance emission for git pushes; a bash validation module that classifies command intent and enforces read-only and permission modes; a bootstrap phase planner for startup sequencing; branch-lock collision detection for worktree safety; session compaction logic that preserves tool-use pairs to prevent API errors; and a comprehensive configuration loader with schema validation, deprecation warning suppression for JSON output, and API timeout/retry settings.
rust/crates/runtime/src · high confidence
New scripts for CC2 board validation, roadmap ID management, and dogfood builds
Added several new scripts to the repository: \cc2\_board.py\ provides a unified wrapper to generate and validate the CC2 board schema; \roadmap-next-id.sh\ and \roadmap-check-ids.sh\ manage unique numeric IDs for ROADMAP entries to prevent collisions; \dogfood-build.sh\ builds the Rust binary with injected provenance (GIT\_SHA) and verifies it; and \fmt.sh\ simplifies running Rust formatting from the repo root.
scripts · high confidence
New telemetry crate for Anthropic request profiling and event tracking
A new \telemetry\ crate has been added to the Rust workspace, introducing core data structures for tracking application behavior and API interactions. It defines \ClientIdentity\ and \AnthropicRequestProfile\ to manage client metadata and automatically generate Anthropic-specific HTTP headers (such as \anthropic-version\ and \anthropic-beta\) and JSON body fields. Additionally, it provides \AnalyticsEvent\ and \SessionTraceRecord\ structs, along with a \TelemetryEvent\ enum, to structure and serialize HTTP request lifecycle events (start, success, failure) for observability purposes.
rust/crates/telemetry · high confidence
Python port introduces session persistence, structured output, and workspace path-scoping
The Python port now supports saving and restoring query sessions (src/session\_store.py, src/query\_engine.py), allowing users to resume previous interactions. The query engine (src/query\_engine.py) adds structured JSON output and token-budget enforcement, while the runtime (src/runtime.py) exposes a bootstrap session that records setup, routing, and execution details. Tool execution (src/tools.py, src/permissions.py, src/path\_scope.py) is now gated by a workspace path-scope validator that denies operations targeting paths outside the configured roots, and a dogfood probe script (scripts/dogfood-probe.py) is added to safely test command-line invocations.
python · high confidence
Behavioural changes
New session metadata file format introduced
A new session metadata file (session-newer.jsonl) has been added to the .claw directory, containing structured session metadata with a version field and timestamps. This represents a change in how session state is persisted or tracked within the CLI tool.
rust/crates/rusty-claude-cli/.claw · low confidence
SDK session lifecycle, reconciliation, and operation registry
The SDK now enforces operation validation via a new OperationRegistry that checks operation names and validates input/output schemas against defined JSON constraints. Session invocations are tracked through a durable reconciliation system that persists invocation state (Accepted, Running, Uncertain, Terminal) to the file system, enabling recovery of in-flight or uncertain requests after process restarts. The transport layer has been upgraded to a reconnecting WebSocket client with configurable timeouts, backoff strategies, and protocol v3 capabilities, while error handling now includes specific codes for connection closure and uncertain delivery states alongside sent-record tracking for request reconciliation. Also in this area: Improved Anthropic credential hints, updated model pricing, and fixed terminal scope path limits; Introduce GJC SDK wire protocol and WebSocket transport.
rust/crates/sdk · high confidence
Subsystem metadata now loads via centralized helper
The \\_\init\\_.py\ files for all subsystems (e.g., assistant, bootstrap, cli) no longer read JSON snapshot files directly using \json\ and \pathlib\. Instead, they delegate to \src.\_archive\_helper.load\_archive\_metadata\, centralizing how archive metadata is retrieved.
(repo-wide) · high confidence
Fixes
Resolve Clippy lint violations in claw-rag-service
Updates the \claw-rag-service\ crate to resolve new Clippy lint violations. In \db.rs\, the \blob\_to\_f32\_vec\ function now uses the stable \as\_chunks\ method instead of \chunks\_exact\ for byte slicing. In \ingest.rs\, the \flush\_path\_batch\ function removes an unnecessary \.into\_iter()\ call when zipping iterators. These changes improve code style and compliance with current Rust lints without altering functional behavior. Also in this area: Support for persisted provider configuration and Groq model routing; Introduce claw-rag-service for workspace RAG indexing and querying.
rust/crates/claw-rag-service · high confidence
Test coverage
Added benchmarks for API request building performance; Added integration and conformance tests for runtime policy and contract validation; Added test coverage for Windows path security and bash detection; Added tests for path scope enforcement; Expanded integration test coverage for CLI flags, config defaults, and output contracts; Integration tests for API client, provider routing, and proxy configuration.
Dependencies
Initial Rust workspace and dependency configuration
The project introduces a new Rust workspace structure, establishing the foundational build system with \Cargo.toml\ and \Cargo.lock\ files. This change defines the workspace members, including core crates such as \api\, \runtime\, \tools\, \plugins\, and the \rusty-claude-cli\ binary, while specifying their external dependencies (e.g., \tokio\, \serde\, \reqwest\, \axum\) and internal path dependencies. It sets the edition to 2021, enforces strict linting rules (forbidding unsafe code), and configures the initial dependency graph required for the Rust-based CLI and agent infrastructure.
(dependencies) · high confidence
Housekeeping
Internal session metadata files added; Ultragoal G010 session hygiene quality gate recorded.
Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.
How this codebase got here
Score
- CAI 58 → 59 (+1.2)
- Rubric changed (rubric-2026.10.5 → rubric-2026.11.4) — scores are not directly comparable.
Lenses
- Code Health 79 → 88 (+8.3)
- Architecture 93 → 93 (-0.0)
- Maturity 65 → 68 (+3.2)
- Readiness 43 → 43 (-0.1)
- Security 68 → 64 (-3.9)
- Event Sourcing 100 → 100 (+0.0)
- Accessibility 78 → 100 (+21.9)
- Performance 91 → 91 (+0.0)
Resolved (42)
- Duplicated block (10 lines × 2) (rust/crates/plugins/src/lib.rs)
- Duplicated block (10 lines × 2) (rust/crates/rusty-claude-cli/src/main.rs)
- Duplicated block (10 lines × 3) (rust/crates/sdk/src/session/reconciliation.rs)
- Duplicated block (10–12 lines × 2) (rust/crates/sdk/src/session/reconciliation.rs)
- Duplicated block (11 lines × 2) (rust/crates/rusty-claude-cli/src/main.rs)
- Duplicated block (11–12 lines × 2) (rust/crates/rusty-claude-cli/src/main.rs)
- Duplicated block (12 lines × 2) (rust/crates/rusty-claude-cli/src/main.rs)
- Duplicated block (12–13 lines × 2) (rust/crates/claw-analog/src/main.rs)
- Duplicated block (12–13 lines × 2) (rust/crates/runtime/src/git_context.rs)
- Duplicated block (15 lines × 2) (rust/crates/commands/src/lib.rs)
- Duplicated block (16 lines × 2) (rust/crates/runtime/src/mcp_stdio.rs)
- Duplicated block (18 lines × 2) (rust/crates/commands/src/lib.rs)
- Duplicated block (21 lines × 2) (rust/crates/runtime/src/session_control.rs)
- Duplicated block (5 lines × 3) (rust/crates/rusty-claude-cli/src/main.rs)
- Duplicated block (6 lines × 2) (rust/crates/rusty-claude-cli/src/main.rs)
- Duplicated block (6 lines × 2) (rust/crates/rusty-claude-cli/src/main.rs)
- Duplicated block (6 lines × 3) (rust/crates/runtime/src/conversation.rs)
- Duplicated block (7 lines × 2) (rust/crates/commands/src/lib.rs)
- Duplicated block (7 lines × 2) (rust/crates/commands/src/lib.rs)
- Duplicated block (7 lines × 2) (rust/crates/commands/src/lib.rs)
- …and 22 more
New (168)
- Duplicated block (10 lines × 2) (rust/crates/plugins/src/lib.rs)
- Duplicated block (10 lines × 2) (rust/crates/plugins/src/lib.rs)
- Duplicated block (10 lines × 2) (rust/crates/rusty-claude-cli/src/main.rs)
- Duplicated block (10 lines × 3) (rust/crates/sdk/src/session/reconciliation.rs)
- Duplicated block (10–12 lines × 2) (rust/crates/sdk/src/session/reconciliation.rs)
- Duplicated block (11 lines × 2) (rust/crates/rusty-claude-cli/src/main.rs)
- Duplicated block (11–12 lines × 2) (rust/crates/rusty-claude-cli/src/main.rs)
- Duplicated block (12 lines × 2) (rust/crates/commands/src/lib.rs)
- Duplicated block (12 lines × 2) (rust/crates/rusty-claude-cli/src/main.rs)
- Duplicated block (12 lines × 2) (rust/crates/rusty-claude-cli/src/main.rs)
- Duplicated block (12 lines × 2) (rust/crates/rusty-claude-cli/src/main.rs)
- Duplicated block (12–13 lines × 2) (rust/crates/claw-analog/src/main.rs)
- Duplicated block (12–13 lines × 2) (rust/crates/runtime/src/git_context.rs)
- Duplicated block (12–14 lines × 2) (rust/crates/rusty-claude-cli/src/main.rs)
- Duplicated block (13 lines × 2) (rust/crates/runtime/src/prompt.rs)
- Duplicated block (16 lines × 2) (rust/crates/runtime/src/mcp_stdio.rs)
- Duplicated block (17 lines × 2) (rust/crates/rusty-claude-cli/src/main.rs)
- Duplicated block (19 lines × 2) (rust/crates/rusty-claude-cli/src/main.rs)
- Duplicated block (22 lines × 2) (rust/crates/runtime/src/session_control.rs)
- Duplicated block (5 lines × 2) (rust/crates/rusty-claude-cli/src/main.rs)
- …and 148 more
Changes since last survey
- 21 commits — 16 feature/other, 5 fixes
By area
- rust/crates — 7 commits
- (repo) — 5 commits
- (root) — 4 commits
- docs/local-openai-compatible-providers.md — 4 commits
- .github/workflows — 1 commit
Notable commits
- fix: fix(api): retry 529 overloaded and gateway-flake 400 responses (#3381)
- fix: fix(runtime): keep tool-use/tool-result pairs paired in Trident compaction (#3367)
- fix: fix(tools): cut RemoteTrigger responses on a char boundary (#3369)
- fix: fix(tools): stop NotebookEdit reusing cell ids after a delete (#3371)
- fix: fix(windows): fix native runtime compatibility and enable workspace t… (#3378)
- change: Merge pull request #3368 from ultraworkers/release/v0.1.9
- change: Merge pull request #3370 from ultraworkers/release/v0.1.10
- change: Merge pull request #3374 from ultraworkers/release/v0.1.11
- change: Merge pull request #3379 from ultraworkers/release/v0.1.12
- change: Merge pull request #3382 from ultraworkers/release/v0.1.13
- change: ci(release): build macOS arm64 on macos-15; macos-14 runners are no longer supported (#3380)
- change: docs: document HopBase OpenAI-compatible gateway setup (#3384)
- change: docs: document LLMTR OpenAI-compatible gateway setup (#3373)
- change: docs: document Standard Compute OpenAI-compatible gateway setup (#3383)
- change: docs: document UberLLM OpenAI-compatible gateway setup (#3376)
- change: release: v0.1.10
- change: release: v0.1.11
- change: release: v0.1.12
- change: release: v0.1.13
- change: release: v0.1.9
- …and 1 more
Architecture
- Unchanged — 0 containers · 1 contexts · 0 edges
Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.
Survey your own repository
ultraworkers/claw-code was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.
About this page
- The score is its most recent published measurement, taken on 14 November 2026 at a pinned commit. It is not a live figure and does not change until the project is measured again.
- Measured at commit 508b4d862d42a5fa5bde8df5b2d3670f4d4ca324 — the exact code this score is about.
- Scored under rubric-2026.11.4 — the same rubric and the same method as every other entry in this index.
- Measured by Canine Code Analyzer (cca.canine.dev) using cca measure-2026.11.65+preprod-36d4cd83012f.