All servers / AI / Realm MCP server
Realm MCP server
MCP server for Realm: lets an AI agent run Realm workflows.
Tested by ConnectATDScore 78 out of 100. Tested 2026-11-19 on package version 0.50.0. Package: @sensigo/realm-mcp · source
Score breakdown
| Runs | 18 / 20 | |
|---|---|---|
| Tool descriptions | 12 / 20 | |
| Package safety | 20 / 20 | |
| Maintenance | 20 / 20 | |
| Adoption | 8 / 20 |
What we saw
- Starts and answers the MCP handshake: yes (5.0 seconds including download)
- Tools offered: 10
- Last published: 2026-11-17
- Weekly downloads (npm): 913
- License: Apache-2.0 · Direct dependencies: 5
Want this server set up for your team?
ConnectATD builds AI automation and wires up MCP servers like this one. Tell us what you are trying to do.
Talk to ConnectATDTools and how they are described
Agents choose tools from these descriptions. Under each name you can see how many of its parameters are explained.
list_workflowsno parametersList all registered Realm workflows; names every registered copy that could not be read, and refuses when the registry directory itself cannot be read.
get_workflow_protocol0 of 1 parameter describedGet the full agent protocol briefing for a registered workflow.
start_run0 of 5 parameters describedCreate a new workflow run and run its first automatic steps.
start_run_batch0 of 6 parameters describedAtomically enqueue multiple runs of the same workflow. No step runs; each started entry's next_actions names its first call. All items are validated before any run is created. If idempotency keys are provided, duplicate runs are returned instead of created.
execute_step0 of 5 parameters describedExecute a workflow step. For agent steps, pass your output in params. Optional: writer_nonce — mint a fresh UUIDv4 per step-attempt (same value on every append_trace call for this attempt, a NEW one next attempt) to get this step's own buffered trace lines faithfully attributed instead of the honest-but-unattributed floor. Mint on BOTH append_trace and execute_step for this attempt, or neither — a guessable or reused nonce is worse than omitting one (it lets residue be adopted with no caveat, or silently folds a crashed attempt into this one).
advance_run0 of 1 parameter describedRun what a run owes the engine — an expired question's declared on_expiry, then its guards and automatic steps — in this server's environment. Call it when next_actions names it.
get_run_state0 of 4 parameters describedGet the current state summary of a workflow run. Pass include_steps: true for each step's model-call cost per attempt, or step_names for the named steps only. When the reply leaves steps out it names next_cursor: pass it as cursor for the next page.
abandon_run0 of 2 parameters describedAbandon a non-terminal workflow run (marks it terminal with phase "abandoned"). Refuses runs waiting on a human gate and runs that have already ended.
create_workflow0 of 3 parameters describedCreate a dynamic Realm workflow at runtime and immediately start a run. Use this when you have a multi-step task and want to track your own execution plan. Returns next_action pointing at your first step — proceed with execute_step as normal. Top-level fields: steps, metadata, and extensions (extensions is register-time, operator-only, and rejected here); any other field is dropped and returned as a warning.
append_trace0 of 4 parameters describedSubmit trace entries incrementally during an agent step, before calling execute_step. Entries are buffered and merged with any entries submitted at execute_step finalization. Delivery is best-effort: entries are durable after this call returns but are not canonical until execute_step completes. Only valid for agent steps that have not yet been claimed (before execute_step is called). Optional: writer_nonce — mint a fresh UUIDv4 per step-attempt (the SAME value on every append_trace call for this attempt, a NEW one next attempt) to get this attempt's own lines faithfully attributed instead of the honest-but-unattributed floor. Mint on BOTH append_trace and execute_step for this attempt, or neither — a guessable or reused nonce is worse than omitting one (it lets residue be adopted with no caveat, or silently folds a crashed attempt into this one).
Tested in a sandbox with an empty environment: no API keys, a throwaway folder, nothing from a real account. A passing test shows the server runs and responds. It is not a security audit.