100 LLM workflow tools for agents, memory, retrieval, and streaming

Toolcore added 100 more client-side helpers for LLM workflow planning—agent and MCP manifest sizing, memory slot budgeting, retrieval keyword stats, multilingual script mix, compliance pattern scans, and streaming/SSE metrics. They live under AI tools alongside earlier token, embedding, and eval batches. None call a generative API.

Agent & MCP manifests

Estimate manifest tokens, count tool schemas, or budget system prompt plus tools—start with MCP manifest token estimate, function definition count, and system prompt + tools budget.

Memory & context

Plan slot char budgets, summary compression ratios, sliding windows, and turn counts—try memory slot budget chars, summary compression ratio, and sliding memory window.

Retrieval & search

Stopword ratio, keyword density, lite TF-IDF, duplicate chunk hashes—see stopword ratio, BM25 TF-IDF lite, and duplicate chunk hash.

i18n & locale

Script mix, RTL counts, bilingual lines, diacritics—open Latin/CJK script mix, RTL char count, and bilingual line ratio.

Compliance & governance

Heuristic email/phone/SSN pattern counts, license headers, API-key shapes—use email pattern count, API key pattern count, and SPDX license detect. These are reports only—not legal or DLP guarantees.

Streaming & SSE

Count SSE data lines, NDJSON rows, tokens/sec samples, WebSocket frame bytes—examples: SSE data line count, tokens per second sample, and NDJSON line count.

Pair with Batch AI-DQ for embeddings and rate limits. Redact secrets with paste redact before sharing exports.