Appearance
MCP Tool Reference
The installed 0.2.0 product exposes 51 MCP tools, including files, experiments, preparation, station selection, instrument Profiles, workflow simulation and desktop views. Available operations still depend on the connected driver.
Using the tools
The AI client discovers and calls these tools. The schema returned by the connected server defines the actual arguments. Users normally work with the natural-language requests on the task pages.
File saves check revisions, plan edits check fingerprints, and run controls and configuration application preserve request identity with request_id. Inspect the original request when its outcome is unresolved. Do not interpret accepted or pending as completed.
Paths are bounded by the service configuration. Ordinary tool calls cannot select an arbitrary draft database. Physical loading confirmation is not exposed as an MCP tool.
Files and Editing
| Tool | Purpose |
|---|---|
file_roots | Read shared workspace and experiment roots. |
list_directory | Read a page of entries in an allowed directory. |
preview_file | Read a bounded text preview or file metadata. |
read_document | Read experiment source and its current revision. |
create_document | Create an experiment without overwriting an existing file. |
save_document | Save against expected_revision; reject stale edits. |
receive_files | Copy a bounded file package into the experiment directory. |
Checking and Execution
| Tool | Purpose |
|---|---|
list_programs | List programs and their source directories. |
preview_program | Preview requirements without creating a run or reserving resources. |
inspect_program | Inspect and freeze a program snapshot. |
start_program | Start a reviewed program; supported simulations advance preparation automatically. |
submit_program | Submit a reviewed snapshot for staged preparation. |
list_runs | List resident runs. |
run_status | Read a run, report and evidence. |
control_run | Request pause, resume or cancellation. |
control_request_status | Read the original control receipt for a run. |
Loading Preparation
| Tool | Purpose |
|---|---|
loading_plan | Read the current loading plan. |
revise_loading_plan | Revise placement preferences against the plan fingerprint. |
begin_loading | Reserve resources and begin preparation; does not confirm physical loading. |
loading_request_status | Read the original loading request receipt. |
Workstation and Simulated Supplies
| Tool | Purpose |
|---|---|
station_status | Read device, station and run observations. |
tip_supply | Read declared simulated tip supplies and replacement history. |
replace_simulated_tip_box | Explicitly change a supported simulated tip load; does not confirm physical replacement. |
tip_replacement_status | Read the original simulated replacement receipt. |
Workstation Selection and Simulation Policy
| Tool | Purpose |
|---|---|
station_selection_status | Read installed choices, saved next-start selection and connected profile. |
select_station_for_next_start | Save a choice against the current selection; does not hot-switch the resident. |
set_workflow_simulation | Change the explicit workflow-model policy against its current revision. |
Restart the App and reconnect MCP when the selection receipt requests it, then verify station_status. Read the workflow policy and revision from station_status before changing it. See simulation modes.
Instrument Profiles
| Tool | Purpose |
|---|---|
instrument_profile_catalogue | Read driver-advertised modules, methods, parameter schemas and defaults. |
list_instrument_profiles | List saved method Profile revisions. |
read_instrument_profile | Read an exact Profile revision. |
validate_instrument_profile | Check proposed method parameters without saving. |
save_instrument_profile | Append a validated revision; check the expected revision when updating. |
Saving does not activate a Profile globally. Supported runs can bind exact revisions per step using profile_bindings on start_program or submit_program. See Instrument Method Profiles.
Configuration and Layout
| Tool | Purpose |
|---|---|
configuration_profiles | Discover installed configuration capabilities and templates. |
list_configurations | List drafts in the shared store. |
read_configuration | Read a draft or an exact saved revision. |
create_configuration | Create a draft without applying it. |
save_configuration | Save a draft against its expected revision. |
validate_configuration | Run static driver checks for an exact revision. |
review_program_bindings | Check a configuration revision against an inspected program snapshot. |
configuration_status | Read active configuration or an application receipt. |
apply_configuration | Apply an exact revision to an eligible idle workstation. |
labware_catalogue | Read editable targets and compatible models for the active configuration. |
prepare_labware_change | Prepare placement, replacement, removal or carrier replacement as a draft. |
import_vendor_layout | Convert a vendor layout to a preview without saving or applying it. |
export_configuration | Return validated configuration JSON. |
export_vendor_layout | Return a device layout archive and integrity information. |
Desktop Views
| Tool | Purpose |
|---|---|
view_state | Read live desktop sessions and reported view state. |
open_view | Request a page in a selected desktop session. |
reveal_entity | Request highlighting of a declared workstation object. |
view_request_status | Read a display receipt; it is not execution evidence. |
Developer Support
| Tool | Purpose |
|---|---|
test_device_fault | Run a supported detached fault test without changing the active station or resuming an experiment. |
Typical sequences
| User task | Sequence |
|---|---|
| Check and start a simulation | Find program → preview / inspect → confirm target → start → run_status |
| Edit a configuration | Read / create → save → validate → confirm revision → apply → configuration_status |
| Change labware | configuration_status → labware_catalogue → prepare → review draft → validate / apply → read back active layout |
| Open a desktop page | view_state → select session → open_view → view_request_status |
A registered tool does not establish support in every driver. Read capabilities and diagnostics before proceeding. Device and configuration queries are routed through the resident's Device Manager. The bridge does not independently discover device plugins; see Connect Your AI Agent.