Thermal label MCP server: render, validate, debug and convert ZPL, EPL, TSPL and CPCL labels.
Scored across 26 tools
The bulk of the set follows a clean per-language matrix (zpl/epl/cpcl/tspl × preview/validate) plus clearly separated resources (barcode, bulk, template, printer, print job), so an agent can almost always pick correctly. The only friction is the cluster of ZPL-checking tools — zpl_validate, explain_zpl and zpl_compatibility all inspect ZPL and rely on their descriptions to separate pass/fail linting from a scored human report from printer-model risk analysis.
Most names are predictable noun_action compounds (bulk_submit, template_get, print_status, zpl_preview) with consistent snake_case and no camelCase drift. A minority invert the order or use verb-first phrasing (explain_zpl, save_label, convert_zpl_dpi, image_to_zpl, printer_language_to_zpl), which slightly breaks the pattern but stays readable and unambiguous.
26 tools is on the heavy side and pushes past the comfortable 3-15 band, largely because four printer languages each get a preview/validate pair. Each tool does map to a genuinely distinct capability rather than a duplicate, but the surface is broad enough that an agent must triage before acting.
Generation, linting, previewing, language conversion, DPI rescaling, bulk production, templates and physical printing are all covered, which is a wide lifecycle. The notable gap is saved labels: save_label creates one, but there is no label_list, label_get, label_update or label_delete, and print jobs can only be polled by id with no listing or cancel operation.
Generate a single barcode as a PNG image (Code 128, QR, DataMatrix, EAN-13, UPC-A, PDF417 and more). For a full label with text and layout, write ZPL and use zpl_preview instead.
Inputs: data (required), type.
Decode barcodes from a photo or scan of a PRINTED label (PNG, JPEG, BMP or GIF) and report what a software decoder reads back. Honest by design: 'read' does not guarantee every handheld scanner will read it, and 'not read' does not condemn the label (photo angle/focus matter) — the response says so.
Inputs: image_base64 (required).
List your recent bulk jobs (id, state, error class, timestamps), newest first. Useful to recover a lost job id; details via bulk_status.
Inputs: limit.
Status of one bulk job you own: state, attempt count and — when finished — the summary with produced count and per-row failures (row number + reason; failed rows never block the rest). Finished outputs stay downloadable for 7 days via the REST endpoint the response names.
Inputs: job_id (required).
Submit a bulk label production job: one ZPL template with {{variable}} placeholders plus data rows (one object per label). Requires an API key whose plan includes bulk jobs. Quota is charged UPFRONT, one operation per row, and failed rows are not refunded; rows per job are capped by plan. Returns the job id — check bulk_status (small jobs usually finish within seconds).
Inputs: zpl (required), rows (required), dpmm, output, +2 more.
Rescale ZPL between printer resolutions (203/300/600 DPI): coordinates, fonts, barcode module width and label size are scaled. Embedded ^GF/~DG image data is NOT scaled; the response says so instead of silently passing it through.
Inputs: zpl (required), source_dpi (required), target_dpi (required).
Render CPCL label code and return a PNG preview. The preview assumes 203 dpi and is an interpretation of the CPCL manual covering a core command subset — no physical printer validation. labels in the response is the TOTAL number of labels the stream prints (a stream may contain several, each ended by PRINT); pass index to render a different one.
Inputs: cpcl (required), index.
Lint CPCL label code and return positioned findings with severity as JSON. Checks the command set Labelixa implements from the CPCL manual — a clean result is not a guarantee that a specific printer accepts the job. Run this before cpcl_preview; if you are not sure which language the code is, call language_detect first.
Inputs: cpcl (required). Returns: summary, diagnostics
Render EPL/EPL2 label code and return a PNG preview. The preview assumes 203 dpi and is an interpretation of the EPL2 manual covering a core command subset — no physical printer validation. labels in the response is the TOTAL number of labels the stream prints (a stream may contain several, each ended by P); pass index to render a different one.
Inputs: epl (required), index.
Lint EPL/EPL2 label code and return positioned findings with severity as JSON. Checks the command set Labelixa implements from the EPL2 manual — a clean result is not a guarantee that a specific printer accepts the job. Run this before epl_preview; if you are not sure which language the code is, call language_detect first.
Inputs: epl (required). Returns: summary, diagnostics
Human-readable health REPORT for a finished ZPL label, with a 0-100 score: sectioned findings (syntax, size/DPI, orientation, barcodes, fonts, graphics memory, job behaviour); sections that cannot be assessed say so and are excluded from the score. Use it to review or explain a label to a person; for a quick pass/fail check while editing, use zpl_validate instead.
Inputs: zpl (required), dpmm, width_in, height_in.
Convert an image (PNG, JPEG, BMP, GIF or SVG, max 200 KB) to a ZPL ^GF graphic command. SVG is rasterised at the chosen density from its declared size (mm/in honoured); SVGs with external references or a DTD are rejected. Returns the ZPL plus measured dimensions and the physical print size at the chosen density. PDF input is REST-only (/v1/graphics).
Inputs: image_base64 (required), dpmm, wrap, dither.
Detect which printer language raw label code is written in (ZPL, EPL, TSPL or CPCL). Heuristic: returns the language plus a confidence TIER (high/medium/low) and signal codes — not a probability.
Inputs: code (required). Returns: mixed, scores, evidence, language, +1 more
Convert a SATO SBPL, Honeywell (Datamax) DPL or Honeywell (Intermec) IPL label stream to ZPL, so it can be checked with zpl_validate and rendered with zpl_preview. You must name the source language: language_detect does not recognise these three. Real control characters and the manuals' readable notation (, ) are both accepted. Commands outside the supported subset (printer settings, RFID, job management, some 2D symbologies) are skipped with a warning, never silently: non-empty warnings mean the conversion is not 1:1 and one_to_one is false — show the warnings to the user. The output follows each manufacturer's command reference and is not verified on physical printers; check a preview and a test print before production.
Inputs: code (required), source_language (required), dpmm, label_length_mm. Returns: zpl, note, warnings, one_to_one, +1 more
List the printers registered to the Labelixa account (the ones a paired Labelixa DirectPrint agent forwards jobs to). Use a returned device id as device_id in print_label. Requires an API key whose plan includes printing.
Inputs: no named parameters.
Send ZPL to a physical printer through the account's Labelixa DirectPrint agent. Requires an API key whose plan includes printing and a printer from printer_list (or a printer group name). The job is QUEUED, not printed: the response says whether the device is mapped by an agent and when the agent last connected; check print_status for delivery. Always show the user a zpl_preview before printing.
Inputs: zpl (required), copies, device_id, printer_group, +1 more.
Status of one print job you own (queued, sent to the agent, delivered or failed) plus when the agent last connected. 'delivered' means the agent handed the bytes to the printer; it does not prove the label came out. printer_status is the printer's own status reading taken after delivery (health ready / blocked / unknown, flags such as paused, head_open, media_out, ribbon_out); null means the printer was not asked, and no reading ever confirms that the label printed.
Inputs: job_id (required).
Save a ZPL label to a Labelixa account so it can be edited, versioned and printed later. With an API key the label is saved immediately and the saved label id is returned. Without a key the label is held for 24 hours and a one-time link is returned: the user opens it, signs in (free account is enough) and the label lands in their account. Nothing else is stored for anonymous calls.
Inputs: zpl (required), dpmm, name, width_in, +1 more.
Fetch one catalog template: metadata, preview image URL and — for non-premium templates — the ZPL source, ready for zpl_preview. Premium templates return metadata for everyone; their ZPL source requires an API key whose plan includes the premium-template feature (same rule as the website).
Inputs: template_id (required).
List Labelixa's first-party label template catalog: id, English name, category, size/density and tags. Read-only product content (no tenant data). Fetch one with template_get; render its ZPL with zpl_preview.
Inputs: search, category.
Render TSPL/TSPL2 label code and return a PNG preview. The preview assumes 203 dpi and is an interpretation of the TSPL manual covering a core command subset — no physical printer validation. labels in the response is the TOTAL number of labels the stream prints (a stream may contain several, each ended by PRINT); pass index to render a different one.
Inputs: tspl (required), index.
Lint TSPL/TSPL2 label code and return positioned findings with severity as JSON. Checks the command set Labelixa implements from the TSPL2 manual — a clean result is not a guarantee that a specific printer accepts the job. Run this before tspl_preview; if you are not sure which language the code is, call language_detect first.
Inputs: tspl (required). Returns: summary, diagnostics
Explain a single ZPL command: syntax, parameters with types/ranges/options, a fragment example and whether the Labelixa renderer draws it. Data comes from the same catalog that powers the public command reference pages.
Inputs: command (required), locale.
Compatibility RISK analysis of ZPL code against a specific printer model, given as manufacturer/model (e.g. 'zebra/zd421'). NOT an emulator and never says 'it works': reports the model's language posture with evidence level and size-rule findings.
Inputs: zpl (required), model (required).
Render ZPL label code and return the label as a PNG image. Use this to SEE what ZPL output looks like — LLMs can write ZPL but cannot otherwise verify the visual result. One call renders one label (use index to pick a label from a multi-label stream). The result also carries an 'Open this label in the Labelixa editor' link the user can click to edit, save or share the label.
Inputs: zpl (required), dpmm, index, width_in, +1 more.
Pass/fail lint for ZPL — machine-readable diagnostics with line/column, code and severity: unknown commands, missing ^FS, out-of-range parameters, fields outside the label. Fast; run it before zpl_preview whenever you generate or edit ZPL. For a human-readable report with a score, use explain_zpl instead.
Inputs: zpl (required), dpmm, width_in, height_in. Returns: rfid, fields, summary, diagnostics
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
MCP server for Api2Pdf — generate PDFs & images from HTML, URLs or office files; merge, barcodes.
HTML and CSS to PDF MCP server with page headers, footers and page numbers. No headless browser.
A remote MCP server to convert HTML documents (URL or raw source) to PDF or screenshots
Hosted MCP server: render any URL to print-stable PDF, PNG, WebP or Markdown
Add one secure layer between your agents and this server.
