Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@OpenFab MCP Lint this FANUC .LS file and flag missing positions or bad motion commands."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
An open-source MCP toolkit for industrial manufacturing workflows, starting with FANUC robot programs and STEP CAD files.
OpenFab MCP v0.1.0 is an offline-first engineering toolkit. It turns common manufacturing engineering files into structured, auditable data and exposes the same capabilities through Python integration functions, a CLI, and an MCP server.
AI coding agents can already work effectively with source code and developer tooling, but traditional manufacturing engineering files and offline robot programs still lack an open, standardized, auditable tool layer for agents.
OpenFab starts with two practical formats:
FANUC .LS robot program source
STEP / STP CAD files
The goal of v0.1.0 is not robot control. It is a small, inspectable offline engineering workflow for parsing, analysis, linting, and draft generation.
Related MCP server: cad-mcp
Codex / Claude / MCP Clients
↓
OpenFab MCP
/ \
FANUC CAD
↓ ↓
.LS files STEP/STPInternally, the CLI and MCP server share the same service boundary:
CLI ─┐
├→ services.py → FANUC / STEP
MCP ─┘See docs/ARCHITECTURE.md for the integration design.
.LSparse .LS source
inspect program structure
preserve source-oriented program semantics without aggressive guessing
lint UFRAME_NUM, UTOOL_NUM, and P[...] references
detect missing and duplicate positions
check J / L motion forms
check basic speed formats
check CNT / FINE termination forms
generate explicitly offline .LS drafts from structured input
STEP schema and selected Part-21 header metadata
topology counts: solids, shells, faces, edges, wires, vertices, compsolids, compounds
bounding box
overall X / Y / Z dimensions
volume when reliably available
center of mass when reliably available
basic B-Rep statistics such as surface area and total edge length
STEP geometry is handled by Open CASCADE Technology through the cadquery-ocp-novtk package. See docs/CAD_STEP_LIMITS.md for explicit limits.
step_analyze
fanuc_parse_ls
fanuc_lint_ls
fanuc_generate_ls_draft
openfab fanuc parse <file.ls> [--json]
openfab fanuc lint <file.ls> [--json]
openfab fanuc generate <draft.json> [-o output.ls] [--json]
openfab step analyze <file.step|file.stp> [--json]
openfab --versionThe MCP stdio entry point is:
openfab-mcpOpenFab requires Python 3.11 or newer.
Base install: MCP server, CLI, and FANUC text functionality. This does not install OCCT/OCP.
pip install -e .Add STEP/STP analysis:
pip install -e ".[step]"Install all OpenFab runtime functionality:
pip install -e ".[all]"For development and tests, install runtime functionality plus development tools:
pip install -e ".[all,dev]"Parse a known-good offline FANUC example:
openfab fanuc parse examples/fanuc/sample_ok.ls --jsonLint the intentionally malformed example:
openfab fanuc lint examples/fanuc/sample_bad.ls --jsonGenerate an offline draft from structured JSON:
openfab fanuc generate examples/fanuc/draft.json -o openfab_draft.lsAnalyze the included 10 × 20 × 30 mm STEP box:
openfab step analyze examples/step/box_10x20x30.step --jsonThe repository examples are synthetic/offline fixtures and are not real production robot programs.
Start the stdio server:
openfab-mcpThe server exposes the same integration functions used by the CLI. MCP protocol tests use the official MCP Python SDK v2 in-memory client (from mcp import Client) against the server object directly.
See docs/API.md for the current data/API notes.
OpenFab MCP v0.1.0 is designed for offline engineering analysis and draft preparation.
It:
does not connect to real FANUC controllers
does not upload programs
does not execute robot motion
does not validate reachability
does not validate collision safety
does not validate DCS or other safety systems
treats generated LS output as an offline draft only
Generated robot programs require review and validation by qualified personnel using appropriate FANUC tooling and simulation before any real-world use.
The release CI installs dependencies in a clean environment and runs the full test suite on Python 3.11, 3.12, and 3.13 on Ubuntu, plus Python 3.12 on a GitHub-hosted macOS ARM64 runner.
Local release checks:
pip install -e ".[all,dev]"
pytest
python -m compileall src
python -m buildOpenFab MCP is licensed under the Apache License 2.0. See LICENSE.
Third-party components retain their own licenses and copyright. See docs/THIRD_PARTY_LICENSES.md.
Generate an explicitly non-production FANUC LS offline draft from structured input.
| Name | Required | Description | Default |
|---|---|---|---|
| draft | Yes |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description carries full behavioral burden. It does disclose that the generated output is explicitly non-production and offline, which are useful safety cues. However, it does not mention side effects, output format, whether files are written, or any validation behavior.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single tight sentence with no filler. It front-loads the action and purpose, though it is slightly terse given how little parameter documentation exists.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
While the output schema exists and may explain return values, the tool has no parameter documentation, no usage context, and no annotations. For a tool accepting a completely open nested object, the description is too minimal to allow an agent to construct a valid invocation confidently.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The schema has one undocumented object parameter with 0% description coverage, so the description must compensate. Saying 'from structured input' adds a little meaning, but it does not explain the shape, required fields, or semantics of the 'draft' object. This is insufficient for an opaque nested object parameter.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description uses a specific verb ('generate') and names both the resource ('FANUC LS offline draft') and the input ('structured input'). The phrase 'explicitly non-production' clarifies intent and differentiates it from parsing/linting siblings.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies this tool is for drafting FANUC LS content before further analysis, parsing, or linting by siblings. However, it does not explicitly state when to prefer this tool over alternatives, nor does it mention any workflow ordering or exclusions.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Lint FANUC LS source offline and return structured diagnostics.
| Name | Required | Description | Default |
|---|---|---|---|
| path | No | ||
| text | No |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden of behavioral disclosure. It does add useful context by stating the operation is offline and returns structured diagnostics, but it does not discuss side effects, permissions, failure modes, or limitations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single, front-loaded sentence where each component adds value: action, resource, mode, and output. There is no redundancy or filler.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
An output schema exists, so omitting return-value details is acceptable, but the description fails to explain how the two optional inputs interact or when each should be used. Combined with zero annotations and zero schema descriptions, this leaves a significant gap in call construction.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 0%, and the description never mentions the path or text parameters, their relationship, or which one should be supplied. The word 'source' is too generic to help an agent construct the correct invocation.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description uses 'Lint' as a specific action and identifies the resource ('FANUC LS source'), the mode ('offline'), and the result ('structured diagnostics'). This clearly distinguishes it from siblings like fanuc_parse_ls and fanuc_generate_ls_draft.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
There is no guidance on when to use this tool versus fanuc_parse_ls or fanuc_generate_ls_draft, nor whether callers should provide path, text, or both. An agent would have to infer the usage context from the tool name alone.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Parse FANUC LS source from exactly one local .ls path or LS text string.
| Name | Required | Description | Default |
|---|---|---|---|
| path | No | ||
| text | No |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden. It does clarify the 'exactly one' exclusivity constraint (path XOR text), which is useful, but it doesn't disclose error behavior for invalid paths, malformed LS syntax, or what happens when both path and text are provided. The expected parse result structure is available in the output schema, which mitigates some gaps.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
One sentence with high information density: the action, resource, and the key input constraint are all present with zero filler. This is appropriately concise for a simple two-parameter tool.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
The output schema is rich and fully defines the return shape (blocks with addresses, operands, etc.), so the description doesn't need to repeat that. However, it doesn't cover error cases, the precise semantics of 'exactly one' enforcement, or any limits on file size/text length. Adequate for a straightforward parser but with room to add failure-mode transparency.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 0%, and the two parameters (path and text) have no per-parameter descriptions. The description does clarify that they are mutually exclusive alternatives, which adds meaning beyond the raw schema, but it doesn't specify path format, text encoding, or precedence rules if both are supplied.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description states a clear verb ('Parse'), a specific resource ('FANUC LS source'), and the input constraint ('exactly one local .ls path or LS text string'). It distinguishes the tool as the parsing entry point compared to sibling tools that format and validate, though it doesn't explicitly name them.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies the tool is for converting raw LS source into structured data, but it doesn't explicitly state when to choose this over the sibling formatting or validation tools, nor does it mention prerequisites or input restrictions beyond 'exactly one' source.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Analyze a local STEP/STP file offline. No CAM, deployment, or robot control is performed.
| Name | Required | Description | Default |
|---|---|---|---|
| path | Yes |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description carries the burden of behavioral disclosure. It discloses that the tool works offline and performs no CAM, deployment, or robot control, which implies a non-interactive, read-only nature. However, it does not explicitly state whether the file is modified or what the analysis returns, leaving some gaps.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is only two sentences, front-loaded with the core action and resource. Every part earns its place, and the exclusion caveat is concise and clear.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a tool with a single path parameter and an output schema, the description is largely complete: it indicates the file type, local scope, offline operation, and what the tool does not do. It does not detail the output content, but the presence of an output schema reduces the need for that in the description.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The schema has zero description coverage for the 'path' parameter, but the description adds that the file is a local STEP/STP file, which clarifies the parameter's meaning. It does not provide additional detail about path format, validation, or extension behavior, so the compensation is only partial.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's function: analyzing a local STEP/STP file offline. The verb 'analyze' and resource 'STEP/STP file' are specific, and the sibling tools are all focused on Fanuc LS files, so this tool is easily distinguishable.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides clear context that this is an offline analysis tool and explicitly states that CAM, deployment, and robot control are not performed. This gives an agent useful exclusion criteria for when to use it, though it does not explicitly name alternative tools.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections. Dates show when Glama detected each change.
v0.1.0fanuc_generate_ls_draftfanuc_lint_lsfanuc_parse_lsstep_analyzestep_analyze is clearly distinct from the FANUC LS tools, and the three fanuc_ tools have reasonably distinct outputs: parse returns structure, lint returns diagnostics, generate produces a draft. fanuc_parse_ls and fanuc_lint_ls both consume LS source, so a small amount of overlap exists.
The FANUC tools follow a consistent fanuc_<verb>_ls pattern, but step_analyze breaks the convention by placing the object before the verb and omitting a domain prefix. The overall naming is readable and mostly predictable.
Four tools is appropriate for this narrow offline manufacturing scope. Each tool serves a distinct purpose: STEP analysis, LS parsing, LS linting, and LS draft generation.
The set covers the core offline workflow for STEP analysis and FANUC LS handling: parse, lint, and generate. Minor gaps exist, such as no direct STEP-to-LS conversion or LS editing utilities, but the explicit offline/non-production scope keeps the surface reasonable.
Unclaimed servers have limited discoverability.
If you are the server author, to access and configure the admin panel.
Lean 4 MCP server: compile, prove theorems, and formalize math with Mathlib.
327 dev tools via REST API and MCP. Generate Dockerfiles, schemas, K8s, APIs, and more.
Self-hosted MCP server: 26 deterministic dev, security, and EVM tools.
MCP Server for Slima - AI Writing IDE for Novel Authors with AI Beta Reader.
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/HB-0921/openfab-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server
