base64
ClaudeEncode UTF-8 text to Base64 (standard or URL-safe) or decode Base64 back to text. Use for Basic-auth headers, data URIs and tokens. Binary payloads are reported rather than shown.
config_to_json
ClaudeParse TOML, INI, .env, Java .properties, XML or JSON Lines text into JSON. Use when the user pastes a config file or log export and needs structured data. XML with a DOCTYPE or entities is rejected for safety; XML attributes appear with an "@_" prefix.
cron_explain
ClaudeDescribe a cron schedule in plain English, for example "/15 9-17 * 1-5". Use when reviewing crontabs, CI schedules or Kubernetes CronJobs. Supports 5-field and Quartz 6/7-field expressions.
csv_to_json
ClaudeParse CSV into JSON. Handles quoted fields, embedded newlines and auto-detects comma, tab, semicolon or pipe separators. Numbers, booleans and null are inferred, but quoted values stay strings. Use when the user pastes CSV or TSV and wants objects.
hash_text
ClaudeCompute an MD5, SHA-1, SHA-256, SHA-384 or SHA-512 digest of text, or an HMAC when a key is given. Use to check checksums or sign webhook payloads. MD5 and SHA-1 are not safe for security.
jmespath_query
ClaudeEvaluate a JMESPath expression (for example "people[?age > 25].name" or "length(items)") against JSON. Use for projections, filters and functions over API responses; the query language is documented at https://jmespath.org. For JSONPath use json_query instead.
json_aggregate
ClaudeGroup an array of objects by one or more fields and compute count, sum, avg, min, max or distinct-count metrics. Numbers are summed exactly and non-numeric values are ignored. Use for totals and breakdowns where mental arithmetic would be error-prone.
json_analyze
ClaudeReport size (pretty vs minified bytes), nesting depth, value-type counts, unique keys, the largest paths, and arrays that contain duplicate items. Use when the user wants to know why a payload is big or where to trim it.
json_canonicalize
ClaudeProduce the RFC 8785 canonical form of JSON (sorted keys, normalised numbers, no whitespace), optionally with its SHA-256 or other digest. Use when signing or hashing JSON, or when two systems must agree on a byte-identical form.
json_diff
ClaudeList every added, removed and changed path between two JSON documents. Use when the user wants to know what differs between two API responses, configs or fixtures. Key order is ignored.
json_escape
ClaudeEscape text so it can sit inside a JSON string (quotes, backslashes, newlines, control characters) or reverse that. Use when embedding text or code in a JSON value, or reading an escaped string from logs.
json_filter
ClaudeKeep the elements of a JSON array that match one or more conditions on fields (dot paths such as "address.city"). Operators: eq, ne, gt, gte, lt, lte, contains, startsWith, endsWith, in, exists. Use for exact, deterministic filtering instead of eyeballing long lists.
json_flatten
ClaudeTurn nested JSON into a single-level object whose keys are paths like "user.address[0].city". Use when loading JSON into a spreadsheet or key-value store, or to compare configs.
json_format
ClaudePretty-print JSON with consistent indentation, optionally sorting keys. Use when the user wants readable JSON or needs to check that text is valid JSON while formatting it. Reports the line and column of any syntax error.
json_lint
ClaudeFind problems that valid JSON can hide: duplicate keys (JSON.parse silently keeps the last), integers beyond 2^53 that lose precision, inconsistent value types across array items, and mixed key naming styles. Use when data parses but behaves strangely.
json_merge
ClaudeDeep merge two JSON objects: nested objects combine, and for other values the right document wins. Choose how arrays combine (replace, concat or union). Use for layering config overrides or combining API responses.
json_minify
ClaudeRemove all whitespace from JSON to make it as small as possible. Use when the user wants compact JSON for an API payload, config value or to save tokens.
json_mock_generate
ClaudeGenerate realistic fake JSON from a JSON Schema (types, enum, min/max, formats such as uuid, email, date-time, uri) or from an example document. Output is deterministic for a given seed. Use for fixtures, demos and tests.
json_patch_apply
ClaudeApply an RFC 6902 JSON Patch (array of operations) or an RFC 7396 JSON Merge Patch (an object, null deletes a key) to a document and return the result. Use when the user has a patch from an API or tool and wants the patched document. Fails with the operation index if a path does not exist.
json_patch_create
ClaudeGenerate an RFC 6902 JSON Patch (add, remove, replace operations) that turns one JSON document into another. Use when the user needs a minimal, machine-applicable change set for an API PATCH request or an audit log.
json_query
ClaudeRun a JSONPath expression (for example "$.store.book[?(@.price < 10)].title") against JSON and return matching values with their paths. Use to extract part of a large response. Filter expressions run in a restricted sandbox.
json_rename_keys
ClaudeRename every key in a JSON document to a naming style (camelCase, PascalCase, snake_case, kebab-case, UPPER_SNAKE_CASE) and/or with an explicit old-to-new mapping, at every depth. Use when adapting an API response to another convention. Fails if two keys would collide.
json_repair
ClaudeExtract and fix JSON from messy text: LLM replies with markdown fences or chatter, single quotes, unquoted keys, trailing commas, comments, Python True/False/None, or truncated output. Use when JSON.parse fails on model output or a hand-edited file.
json_schema_fix_strict
ClaudeRewrite a JSON Schema so it passes OpenAI Structured Outputs strict mode: additionalProperties:false everywhere, every property required (optional ones become nullable), and unsupported keywords removed. Use when a tool or response_format schema is rejected. Returns the fixed schema and a change list.
json_schema_generate
ClaudeInfer a JSON Schema (draft 2020-12) from a sample document: types, required keys, and string formats such as date-time, email, uuid and uri. Arrays of objects are merged so keys missing from some items become optional. Use when the user needs a schema for validation or an API contract.
json_schema_validate
ClaudeValidate a JSON document against a JSON Schema (draft-07 or 2020-12) and list every error with its path. Use when the user asks if data matches a schema. Schemas using "pattern" or remote $ref are rejected on this server for safety.
json_sort_keys
ClaudeSort the keys of every object in a JSON document, ascending or descending. Use for stable diffs, canonical-looking config files or reproducible output.
json_split_array
ClaudeSplit a JSON array into batches of a fixed size. Use when an API or model accepts only N items at a time.
json_to_code
ClaudeGenerate type definitions from a sample JSON document: TypeScript interfaces, Zod schemas, Pydantic models, Go structs, Rust serde structs, Java records, C# classes, Kotlin data classes, Swift Codable structs or Dart classes. Nested objects become named types and keys missing from some array items become optional. Use when the user needs models for an API response.
json_to_config
ClaudeWrite JSON as TOML, INI, .env (nested keys become UPPER_SNAKE names), Java .properties (dotted keys) or JSON Lines (one array element per line). Use when the user needs a config file or log-style output from JSON. TOML cannot represent null.
json_to_csv
ClaudeConvert an array of JSON objects (or one object) to CSV. Nested objects become dotted columns and quoting follows RFC 4180. Use when the user wants spreadsheet-ready data from an API response.
json_to_definition
ClaudeTurn a sample JSON document into a GraphQL schema, Protobuf (proto3) messages, SQL CREATE TABLE plus INSERT statements (PostgreSQL, MySQL or SQLite), an Avro record schema, an OpenAPI 3 components schema, or a Prisma model. Use when the user has data and needs the matching definition.
json_to_table
ClaudeRender an array of JSON objects as a Markdown, HTML or reStructuredText table. Nested objects become dotted columns. Use when the user wants JSON data in a README, docs page or email.
json_to_toon
ClaudeEncode JSON as TOON (Token-Oriented Object Notation), a compact format that cuts LLM token use for uniform arrays. Use when the user wants to shrink JSON going into a prompt.
json_to_xml
ClaudeConvert JSON to indented XML. Arrays repeat the element, names that are not valid XML are sanitised and text is escaped. Use when a SOAP or legacy system needs XML.
json_to_yaml
ClaudeConvert JSON to YAML. Use for Kubernetes, CI or config files that must be YAML.
json_truncate
ClaudeShorten JSON for display or prompts by limiting nesting depth, array length and string length. Cut arrays end with "… N more items". Use when a document is too big to show in full.
json_unflatten
ClaudeRebuild nested JSON from a flat object of path keys such as "a.b[0].c". Use to reverse json_flatten or to expand dotted environment-style keys.
json_validate
ClaudeCheck whether text is valid JSON and describe it (type, depth, size, key count). Use when the user asks if something parses, or wants the exact line and column of a syntax error. Returns a result even for invalid input.
jwt_decode
ClaudeDecode the header and payload of a JSON Web Token and show expiry, issue time and warnings such as alg "none". Use when debugging authentication. The signature is NOT verified, so never use the result to trust a token.
llm_tool_schema
ClaudeCreate a function-calling tool definition for Claude (input_schema), OpenAI (function, optional strict mode), Gemini (functionDeclarations) or MCP (inputSchema) from a name, description and parameters. Parameters can be a JSON Schema or an example arguments object to infer from. Use when wiring a tool into an agent.
mcp_config_generate
ClaudeWrite the configuration to add an MCP server to Claude Code (command and .mcp.json), Claude Desktop, Cursor or VS Code, for a remote https URL or a local command. Use when the user wants to install or share an MCP server.
qr_code_svg
ClaudeCreate a QR code for text or a URL and return it as SVG markup that scales to any size. Use when the user needs a QR code for a link, Wi-Fi string or contact card.
sql_format
ClaudeFormat SQL for readability in a specific dialect (PostgreSQL, MySQL, SQLite, BigQuery, Snowflake, T-SQL and more). Use when the user pastes a long one-line query.
text_diff
ClaudeCompare two texts line by line and return a unified patch with added and removed line counts. Use for comparing config files, prompts or code snippets.
text_stats
ClaudeCount words, characters, sentences, paragraphs and lines, and estimate reading and speaking time and the most frequent words. Use when the user needs exact counts, which language models routinely get wrong.
timestamp_convert
ClaudeConvert between epoch seconds, milliseconds, ISO 8601 and readable dates, optionally showing local time in an IANA time zone. Use when reading log timestamps or checking token expiry. Accepts "now".
token_count
ClaudeCount tokens with the o200k_base tokenizer (GPT-4o family; a close estimate for other models). With compareFormats, JSON input is also counted as pretty JSON, minified JSON, YAML and TOON to show which saves the most tokens. Use when budgeting prompt or context size.
url_encode
ClaudePercent-encode text for use in a URL or decode it. Use when building query strings or links: scope "component" for a query value or path segment, "url" to keep a whole URL intact.
uuid_generate
ClaudeGenerate random identifiers: UUID v4, time-ordered UUID v7, or ULID. Use when the user needs test IDs, primary keys or fixtures. Returns one per line; each call gives new values.
yaml_to_json
ClaudeConvert a single YAML document to JSON. Use when the user has YAML config and needs JSON. Multi-document files and alias bombs are rejected.