Anthropic SDK
AI & LLMsStandardized changelog data for anthropic-sdk.
Release History
View all versions →v1.2.0Breaking6 fixes1 featureThis release stabilizes the files/skills namespaces by adopting GA shapes and dropping dated beta header pins. It also includes several bug fixes, notably for binary file uploads via AWS/Bedrock and preserving file bytes in toolsets.
v1.1.02 fixes3 featuresThis release introduces new API features including the 'updates' thinking display mode and support for Organization API endpoints. It also includes bug fixes for tool runner behavior and documentation links.
v1.0.0Breaking1 fix1 featureThis release upgrades the client to httpx2, introducing minor breaking changes and bug fixes. Consult MIGRATION.md for detailed upgrade instructions.
v0.125.01 featureIntroduced new features for managed agents, including web search configuration and self-hosted sandbox memory.
v0.124.02 featuresThis release makes the Files and Skills APIs generally available and introduces new toolsets for computer and browser use.
v0.123.06 fixes3 featuresThis release introduces new features for file and memory store management, along with improvements to client headers and error handling. It also includes several bug fixes, such as retrying tool-result sends and running synchronous session tools in a worker thread.
v0.122.014 fixes1 featureThis release introduces `output_behavior` for dream creation and includes numerous bug fixes across API clients, streaming, and tool handling. It also refactors the use of the `NOT_GIVEN` alias to the `not_given` singleton.
v0.121.02 featuresThis release introduces the `mid-conversation-tool-changes-2026-07-01` beta API and adds support for session budgets, advisor tool, pinned inference location, and skills auto-loading from GitHub. It also removes retired Claude Opus 4.1 models.
v0.120.21 fixThis release includes a bug fix to support MCP SDK v2 alongside v1. The changelog provides a link to the full comparison between versions.
v0.120.11 fixThis patch release addresses a bug by pinning the 'mcp' extra to a version less than 2, ensuring better dependency management.
v0.120.03 featuresThis release introduces the new claude-opus-5 model, enhances tool management with addition/removal blocks and tool_change events, and expands credit token handling with client-side fallbacks and server-side defaults.
v0.119.01 fix1 featureThis release introduces a new stop reason for exceeding the model's context window and fixes an issue with handling binary files in the agent toolset.
v0.118.03 featuresThis release introduces new API capabilities including support for Managed Agents model effort, initial session events, and threads delta streaming.
v0.117.11 fixThis release includes a bug fix for AWS credential handling and several dependency updates and documentation improvements.
v0.117.01 fix2 featuresThis release introduces support for dreaming and MCP Tunnels, and includes a bug fix to improve credential security in tracebacks.
v0.116.01 featureThis release introduces support for the new agent-memory-2026-07-22 beta API header.
v0.115.1This patch release primarily focuses on maintenance by removing some nonfunctional types from the SDKs.
v0.115.05 featuresThis release introduces significant new API capabilities, including support for Managed Agents event delta streaming, agent overrides, reverse pagination, and webhook events.
v0.114.01 fix1 featureThis release introduces support for the new claude-sonnet-5 model and fixes a bug related to absolute path handling in the agent toolset.
v0.113.01 fix2 featuresThis release introduces support for the 20260318 web fetch API and tools, alongside a fix for an async count_tokens issue.
v0.112.01 fix1 featureThis release introduces support for streaming system.message events and fixes a bug related to directory permissions in the memory tool.
v0.111.01 featureRelease 0.111.0 introduces a new feature for tagging refusal-fallback middleware requests within the helpers module.
v0.110.03 fixes1 featureVersion 0.110.0 introduces support for the new code_execution_20260120 tool and includes several fixes related to header merging and Bedrock streaming.
v0.109.2This patch release primarily involved chore updates, specifically removing retired models from the API and SDKs.
v0.109.11 fixThis patch release primarily addresses an API update by adding support for the new `frontier_llm` refusal category.
Common Errors
BadRequestError15 reportsBadRequestError in the anthropic-sdk usually arises from sending malformed requests to the Claude API, such as including unexpected or incorrectly formatted data fields like extra headers. To resolve this, carefully inspect your request payload, especially custom headers, output format specifications, and any extra fields, ensuring they strictly adhere to the API's requirements and documented schema; remove any extraneous or incorrect elements that are not explicitly supported.
InternalServerError5 reports"InternalServerError" in anthropic-sdk often indicates a problem on Anthropic's servers, or that the request sent by the sdk encountered an issue like malformed data. Ensure your input data (especially `input_examples`) is correctly formatted according to the API documentation, and implement retry logic with exponential backoff to handle temporary server-side issues. If the error persists despite valid inputs and retries, report the issue to Anthropic with detailed logs and request payloads.
NotFoundError4 reportsThe "NotFoundError" in anthropic-sdk often arises from incorrect package installation or version conflicts, preventing the program from locating necessary modules or attributes. Ensure that the anthropic-sdk is correctly installed (`pip install anthropic --upgrade`) and that its version aligns with the code's requirements, checking for potential conflicts with other installed packages. Downgrading or upgrading anthropic-sdk to a compatible version might also be necessary.
RateLimitError3 reportsRateLimitError in anthropic-sdk usually means you've exceeded your API usage limits. To fix this, check your Anthropic account to understand your rate limits and usage, then implement retry logic with exponential backoff in your code to gracefully handle rate limiting.
PermissionDeniedError2 reportsThe PermissionDeniedError in anthropic-sdk usually stems from insufficient permissions to access the required files or resources, often during installation. Try running the installation with elevated privileges (e.g., using `sudo pip install anthropic` or installing within a virtual environment) to grant the necessary access rights. You may also need to check and correct the permissions of the relevant directories involved in the installation process.
AttributeError2 reportsRelated AI & LLMs Packages
AutoGPT is the vision of accessible AI for everyone, to use and to build on. Our mission is to provide the tools, so that you can focus on what matters.
Get up and running with OpenAI gpt-oss, DeepSeek-R1, Gemma 3 and other models.
🦜🔗 The platform for reliable agents.
The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface.
LLM inference in C/C++
GPT4All: Run Local LLMs on Any Device. Open-source and available for commercial use.
Subscribe to Updates
Get notified when new versions are released