Skip to main content

Tool catalog

This is the full list of registry tools exposed over the NexSpace MCP surface — 100 tools across nine categories. Use it to work out what a scope grant actually unlocks before you connect, and to review the surface you are about to hand to an agent. You need this list because tools/list is not filtered by your credential’s scopes: it advertises the whole governed catalog regardless of what you were granted, so it cannot tell you what scheduling:write or crm:write will unlock. See What tools/list actually returns.
These are the registry tools only. The 19 named workforce verbs (fillShift, runPayroll, searchStaff, …) are a separate layer with their own scopes — see MCP → Named workforce verbs. Connected-app actions are a third layer, gated on apps:read / apps:write and listed per caller. Five in-app tools are deliberately withheld from MCP entirely; see Not available over MCP.

How scopes are derived

Every registry tool’s required scope is mechanical:
  • a read-only tool requires <category>:read
  • any other tool requires <category>:write
There is no per-tool scope and no read/write aliasing — the dispatcher matches the scope string literally (or via a <resource>:* wildcard). Getting the category right is the whole game.
Write tools in read-only categories are unreachable. Only three categories are promoted to read+write: communications, crm, and scheduling. In the other six, a write tool has no headless scope at all — it is not listed, and calling it by name returns -32004 NOT_FOUND (“Unknown tool: …”).The clearest example is compliance’s run_batch_exclusion_check. It exists in the in-app assistant, it is a genuine write, and compliance:read will never unlock it — there is no compliance:write scope to grant. Do not plan an integration around a writer in analytics, calendar, compliance, facility, general, or marketing.

Category summary

analytics

19 read-only tools. All require analytics:read.

calendar

4 read-only tools. All require calendar:read.

communications

Read + write. 1 read tool and 6 write tools. Writes (each returns pending_approval when governance requires sign-off):

compliance

4 read-only tools. All require compliance:read. The category is read-only, so its write tools (including run_batch_exclusion_check) are unreachable over MCP.

crm

Read + write. 13 read tools and 7 write tools. Writes:
The legacy update_deal_stage is not on this surface — use move_deal_stage. Note also that the named verbs searchLeads and qualifyLead require the literal scope crm:*, which crm:read does not satisfy. See the crm:* trap.

facility

3 read-only tools. All require facility:read — not facilities:read, which is the verb scope for listFacilities / getFacility.

general

7 read-only tools. All require general:read.
get_approval_status is how you follow up on a pending_approval write, and it is gated on general:read. A credential minted with only write scopes cannot poll its own approvals. Request general:read alongside any write scope.

marketing

12 read-only tools. All require marketing:read.

scheduling

Read + write. 7 read tools and 17 write tools — the largest category, and the one whose :write grant covers the most surface. Shift writers: Schedule-settings writers:
high-risk writes sit at the un-relaxable risk floor: whenever the call is authorized at all they return status: "pending_approval" and never commit inline, on any surface. Build these paths around the approval round-trip. See What decides whether a write pauses.

How this list is maintained

The exposed set is pinned in CI. A newly registered tool in an exposed category fails the build until someone classifies it — either adding it to the pin or adding it to the denylist — so this page cannot drift silently, and the surface cannot grow by accident.
A live tools/list on your own credential is still the runtime source of truth. It reflects tools promoted since this page was published and, unlike this page, includes your connected-app tools. Use this catalog to plan scopes; use tools/list to drive a client.

See also

  • MCP — the named verbs, the four tool layers, and the error model.
  • Tenant isolation — how facilityId is rewritten or rejected on every tool here.
  • Governance — why a write may return pending_approval instead of a result.
  • Model tier gate — why a tool you are scoped for can still return -32003.