Build reliable AI workflows,
one clear step at a time.
Practical local tools for prompt design, structured output, retrieval context, API configuration, and text cleanup. Choose a workflow below and keep your inputs in your browser.
Prompt design & iteration
Build structured prompts, fill reusable templates, expand batches, and compare revisions.
Prompt Diff & Version Compare
Compare prompt revisions by line, word, or grapheme with real token deltas and optional local versions.
Structured Prompt Builder
Build deterministic prompts from typed fields, CO-STAR or RTF structures, ordered examples, and explicit local templates.
Prompt Batch Variable Expander
Expand {{variable}} prompts from JSON or CSV records with warnings, paginated previews, and lossless exports.
Prompt Template Formatter
Fill {{variable}} placeholders from JSON and identify any values your prompt still needs.
Structured output reliability
Repair malformed model JSON and validate the result against explicit Schema constraints.
RAG, tokens & context
Prepare retrieval chunks and understand token usage, context capacity, and request cost.
Tokenizer Playground
See exactly how text becomes tokens, then compare the cost across popular LLMs.
Token Counter
Count characters and LLM tokens with a lightweight, private text workspace.
Token Cost Calculator
Estimate input and output spend with transparent per-million-token pricing.
RAG Chunking Playground
Preview RAG chunks with six strategies, real token counts, source offsets, and exportable results.
Context Window Budget
Measure prompt tokens and calculate the context left after reserving space for model output.
Model prices & context
Compare model prices, context windows, tokenizer mappings, and sourced cost data.
API request configuration
Assemble provider-native request bodies and inspectable code examples without credentials.
Text cleaning & inspection
Expose invisible Unicode and make selective, auditable cleanup changes locally.
OpenAI models use verified tiktoken mappings. Other providers are labeled as estimates because their production tokenizers differ. Costs use build-time price snapshots, never runtime API calls.