Burs-IA Utils

ONLINE Burs-IA Utils answered an MCP connection on 01 Nov 2026 and introduced itself as bursia-utils 1.5.0 (MCP protocol 2025-06-18).

50+ tools offered to AI assistants

• find_tool · Searches the 465 tools of this server by keywords (English or Spanish) and an optional family, and returns the best matches with their input schemas. When: the right tool for a task is not known yet. Not for: running a tool (tools/call does
• token-estimate · Heuristically estimates how many LLM tokens a text uses (Latin text ~4 chars/token, CJK ~1 char/token, code and punctuation denser) with a ±25% range. Real counts depend on each model's tokenizer. When: an agent loop needs an exact, repeata
• prompt-template-fill · Fills {{variable}} placeholders (dotted paths allowed, optional {{name|default}}) from a values object, reporting missing and unused variables; strict mode fails on missing values. No code execution. When: an agent loop needs an exact, repe
• json-repair · Repairs almost-JSON typical of LLM output (code fences, comments, single quotes, unquoted keys, trailing or missing commas, Python True/False/None, NaN, unclosed strings and brackets) and returns the parsed value and the fixes applied. When
• json-extract · Finds and parses JSON objects/arrays embedded in free text (e.g. an LLM reply): fenced ```json blocks first, then balanced {…}/[…] spans; optionally repairs near-JSON. When: an agent loop needs an exact, repeatable result for a bookkeeping
• code-blocks-extract · Extracts fenced code blocks (``` or ~~~) from Markdown or LLM output with their language tags and line counts, optionally filtered by language. When: an agent loop needs an exact, repeatable result for a bookkeeping step (tokens, costs, ret
• rag-chunk · Splits text into chunks of at most N characters or estimated tokens with overlap, preferring paragraph, then sentence, then word boundaries; returns offsets for citation. When: an agent loop needs an exact, repeatable result for a bookkeepi
• dedupe-similar · Groups near-duplicate strings (Jaccard similarity of character 3-grams ≥ threshold, accent- and case-insensitive) among up to 2,000 items, keeping the first of each group. When: an agent loop needs an exact, repeatable result for a bookkeep
• vector-similarity · Computes the cosine similarity, dot product, Euclidean and Manhattan distance between two numeric vectors of equal length (≤ 8,192 dims), plus their L2 norms. When: an agent loop needs an exact, repeatable result for a bookkeeping step (tok
• vector-top-k · Ranks up to 2,000 candidate vectors by cosine similarity (or negative Euclidean distance) to a query vector and returns the top k with indices and scores; total size capped at 2M numbers. When: an agent loop needs an exact, repeatable resul
• bm25-rank · Ranks up to 2,000 documents for a query with Okapi BM25 (k1, b tunable) using accent-insensitive Unicode word tokens, e.g. to pick passages for retrieval without embeddings. When: an agent loop needs an exact, repeatable result for a bookke
• backoff-schedule · Computes retry delays for exponential, linear or constant backoff with a cap and optional jitter (full, equal or decorrelated; seeded for reproducibility), and the cumulative wait. When: an agent loop needs an exact, repeatable result for a

What "online" means: the server answered an MCP connection at the time of the check. It is not a guarantee of security, quality or continuous uptime. How we test: we open an MCP session and ask for the tool list, as an AI assistant would. We never run a tool and never send a key. Checked every few days. Method

Is this your server? Monitoring Pro: a check every 30 minutes and an email within the hour when it stops answering. Launch offer: $39.99 once, for life.