The AI Toolchain — issue -192, February 6, 2026

Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflows - all through natural language commands.

Claude Code v2.1.33 adds multi-agent hook events, sub-agent spawn restrictions, persistent agent memory, and VSCode remote sessions.

$ git clone --branch v2.1.33 https://github.com/anthropics/claude-code.git
# already have the repo? check out this version:
$ git checkout v2.1.33

# In agent frontmatter (tools section):
---
tools:
  - Task(researcher)
  - Task(summarizer)
---

# In agent frontmatter:
---
memory: project
---

• ›Adds TeammateIdle and TaskCompleted hook events for multi-agent workflows, enabling automation when agents become idle or complete tasks.
• ›Supports restricting which sub-agents can be spawned via Task(agent_type) syntax in agent "tools" frontmatter.
• ›Adds memory frontmatter field for agents with user, project, or local scope for persistent memory across sessions.
• ›Adds plugin name to skill descriptions and the /skills menu for improved discoverability.
• ›Improves API connection failure error messages to show specific causes (e.g., ECONNREFUSED, SSL errors) instead of generic messages.

AI agent toolkit: unified LLM API, agent loop, TUI, coding agent CLI

Pi v0.52.7 adds per-model config overrides and unauthenticated Bedrock proxy support via new env vars.

$ git clone --branch v0.52.7 https://github.com/earendil-works/pi.git
# already have the repo? check out this version:
$ git checkout v0.52.7

# In models.json
{
  "providers": [...],
  "modelOverrides": {
    "claude-opus-4-5": {
      "maxTokens": 8192
    }
  }
}

• ›Adds modelOverrides key in models.json to customize individual built-in provider models (e.g., context window, parameters) without replacing the entire provider model list.
• ›Supports merge-by-id behavior for models.json provider models, letting custom model entries upsert or extend built-ins rather than replacing the full list.
• ›Supports unauthenticated Bedrock proxy endpoints via new AWS_BEDROCK_SKIP_AUTH and AWS_BEDROCK_FORCE_HTTP1 environment variables.

Qwen Code v0.10.0 adds session resume, session history export, Agent Skills GA, Java SDK alpha, and large-paste UI improvements.

$ git clone --branch v0.10.0-preview.1 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.10.0-preview.1

• ›Adds export command to export session history in Markdown and HTML formats.
• ›Adds resume and continue options to the SDK, plus extended authType support via feat(sdk) changes.
• ›Adds FORK_MODE support to ProcessTransport for Electron IPC integration.
• ›Adds support for resuming sessions by session ID in the query command.
• ›Promotes Agent Skills from experimental to stable.

$ git clone --branch v0.10.0-preview.0 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.10.0-preview.0

LangChain 1.2.9 adds state updates from wrap_model_call and threading context through create_agent flows and middleware.

$ git clone --branch langchain==1.2.9 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout langchain==1.2.9

• ›Supports state updates from wrap_model_call with commands, enabling middleware to propagate state changes back through the call graph.
• ›Threads context through create_agent flows and middleware, making request-scoped context available across agent creation and middleware layers.

Build resilient agents.

LangGraph Python SDK gains cron job update, enable, and disable methods in the crons client.

$ git clone --branch sdk==0.3.4 https://github.com/langchain-ai/langgraph.git
# already have the repo? check out this version:
$ git checkout sdk==0.3.4

• ›Adds update method to the crons client for modifying existing cron jobs.
• ›Supports enabling and disabling cron jobs via the crons client.

OpenAI Agents SDK v0.8.1 adds run-context thread reuse for codex_tool and a max-turns limit for the REPL.

$ git clone --branch v0.8.1 https://github.com/openai/openai-agents-python.git
# already have the repo? check out this version:
$ git checkout v0.8.1

• ›Adds a max-turns limit to the REPL loop, preventing runaway multi-turn agent sessions.
• ›Adds run-context thread reuse for codex_tool, allowing tool invocations within a run to share execution context across turns.

PydanticAI v1.56.0 adds Claude Opus 4.6 support, adaptive thinking, and new Anthropic model settings fields.

$ git clone --branch v1.56.0 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:
$ git checkout v1.56.0

from pydantic_ai import Agent
from pydantic_ai.models.anthropic import AnthropicModel, AnthropicModelSettings

agent = Agent(
    AnthropicModel('claude-opus-4-6'),
    model_settings=AnthropicModelSettings(
        anthropic_effort='auto',
        anthropic_thinking={'type': 'adaptive'}
    )
)
result = agent.run_sync('Explain quantum entanglement.')
print(result.output)

from pydantic_ai import Agent
from pydantic_ai.models.anthropic import AnthropicModel, AnthropicModelSettings

agent = Agent(
    AnthropicModel('claude-opus-4-6'),
    model_settings=AnthropicModelSettings(
        anthropic_betas=['interleaved-thinking-2025-05-14']
    )
)
result = agent.run_sync('Draft a threat model for a SaaS API.')
print(result.output)

• ›Adds anthropic_effort and anthropic_thinking.type='adaptive' to Anthropic model settings, enabling adaptive extended thinking for Claude models.
• ›Adds anthropic_betas field to AnthropicModelSettings, allowing opt-in to Anthropic beta features per request.
• ›Adds support for Claude Opus 4.6 as a new model option.

browser-use 0.11.9 adds agent planning, loop detection, message compaction, HAR capture, and a new read_long_content action

$ git clone --branch 0.11.9 https://github.com/browser-use/browser-use.git
# already have the repo? check out this version:
$ git checkout 0.11.9

• ›Adds read_long_content action for agents to handle long webpages and files that exceed normal context limits.
• ›Adds HarRecordingWatchdog for HTTPS HAR capture, enabling recording of network traffic in HAR format over HTTPS.
• ›Introduces basic planning to agents, allowing the agent to reason about a plan before executing steps.
• ›Adds action loop detection with a nudge mechanism to break agents out of repetitive action cycles.
• ›Adds message compaction for agents to manage long conversation histories and stay within context limits.

ONNX Runtime v1.24.1 adds plugin EP APIs, free-threaded Python 3.13t/3.14t, CUDA Graph on by default, and broad new op coverage across CUDA, QNN, WebGPU, and CPU.

$ git clone --branch v1.24.1 https://github.com/microsoft/onnxruntime.git
# already have the repo? check out this version:
$ git checkout v1.24.1

$ ./build.sh --config Release --enable_arm_neon_nchwc

• ›Adds OrtApi::CreateEnvWithOptions() and OrtEpApi::GetEnvConfigEntries() for programmatic environment configuration in plugin execution providers.
• ›Adds OrtKernelInfo APIs for kernel-based plugin EPs, enabling kernel authors to query session configuration from KernelInfo.
• ›Adds add_external_initializers_from_files to the Python binding for loading external initializers directly from file paths.
• ›Adds --enable_arm_neon_nchwc build flag to enable NCHWc layout support for Conv models on ARM NEON, improving convolutional inference performance.
• ›Enables CUDA Graph by default in the CUDA EP, reducing per-inference launch overhead without additional configuration.

llama.cpp b7957 adds Kimi-Linear model support with backend-agnostic inference and MLA KV cache.

$ git clone --branch b7957 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b7957

• ›Adds support for the Kimi-Linear (Kimi-K2) model architecture, including convert_hf_to_gguf.py conversion, backend-agnostic inference, and MLA KV cache via the llm_arch_is_hybrid path.
• ›Introduces llm_graph_input_mem_hybrid_k class to support MLA-style KV cache for hybrid architectures like Kimi-Linear.
• ›Adds Kimi-K2-specific end-of-generation (EOG) tokens to the vocabulary recognition layer.
• ›Adds causal_conv1d static function abstracting convolution logic for query, key, and value projections in Kimi-Linear layers.
• ›Supports Kimi-Linear MoE parameters including n_experts, n_experts_used, and score_func via the shared TextModel conversion path in convert_hf_to_gguf.py.

Arize Phoenix 12.34.0 adds a tool_selection evaluator to the evals library.

$ git clone --branch arize-phoenix-v12.34.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v12.34.0

• ›Adds tool_selection evaluator to the evals library for assessing whether an LLM chose the correct tool during a tool-use span.

Langfuse v3.151.0 adds dataset versioning, a prompts filter-options endpoint, and a parentObservationId null filter for the observations API.

$ git clone --branch v3.151.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v3.151.0

$ curl -X GET 'https://<your-langfuse-host>/api/public/prompts/filterOptions' \
  -H 'Authorization: Basic <base64-encoded-key>'

• ›Adds GET /api/public/prompts/filterOptions endpoint to retrieve filterable metadata for prompts.
• ›Adds parentObservationId null filter to the observations public API, enabling queries for root-level observations only.
• ›Implements dataset versioning support across APIs and schemas, with version resolution for experiment runs in the UI.
• ›Introduces preferredClickhouseService: EventsReadOnly configuration option for routing reads to an events-only ClickHouse replica.
• ›Adds a v4 event-based session view in the UI.

LanceDB v0.26.0 adds VoyageAI v4 embeddings, exposes fast_search in the sync Python API, and introduces storage options APIs.

$ git clone --branch v0.26.0 https://github.com/lancedb/lancedb.git
# already have the repo? check out this version:
$ git checkout v0.26.0

$ results = table.search(query_vector).fast_search().limit(10).to_list()

• ›Exposes fast_search in the synchronous Python API, enabling approximate nearest-neighbor search without switching to the async client.
• ›Adds initial and latest storage options APIs for configuring object-store settings at table open/create time.
• ›Allows the permutation builder memory limit to be configured via environment variable.
• ›Adds VoyageAI v4 embedding models to the Python integration.
• ›Implements TableProvider::insert_into() for LanceDB tables in the Rust API, enabling DataFusion-native inserts.

$ git clone --branch python-v0.29.0 https://github.com/lancedb/lancedb.git
# already have the repo? check out this version:
$ git checkout python-v0.29.0

results = table.search(query_vector).fast_search().to_list()