Read the programme record
Search and retrieve requirements, systems, interfaces, risks, changes, tests and evidence using the access already granted to the Arc user.
Model Context Protocol
Arc’s MCP server lets compatible AI clients search the engineering graph, follow traceability, assess change impact and prepare controlled updates. Permissions and human review stay in Arc.
Explore MCP capabilities
Use natural-language tools without flattening the programme into a chat transcript. Arc returns the records, relationships and controlled actions behind the answer.
Search and retrieve requirements, systems, interfaces, risks, changes, tests and evidence using the access already granted to the Arc user.
Move upstream to intent and downstream to allocations, interfaces, verification work and evidence without rebuilding the trace from separate files.
Ask which connected records may be affected, inspect the cited paths and turn the result into a reviewable set of engineering actions.
Create or update records, connect relationships and prepare changes on a branch. Reviewers remain responsible for approval and merge.
This illustrative CubeSat power example uses explicit records and arithmetic. It demonstrates the control flow, not a customer result or flight design.
Download the example datasetA user asks a compatible AI client what would be affected by changing PAY-PWR-014 from an 18 W design point to 24 W.
The client uses Arc MCP to retrieve the current requirement revision, its 20 W cap, linked electrical interface and verification plan.
The connected example moves total peak demand from 54 W to 60 W against 59 W available, changing margin from +5 W to −1 W.
The response cites the affected records, states the 4 W cap exceedance and identifies the interface and verification work that need review.
If authorised, the client prepares a branch with proposed record updates and review actions. It does not silently rewrite the accepted requirement.
An engineer inspects the diff, assumptions and evidence, then approves, revises or rejects the proposed change in Arc.
MCP does not bypass Arc’s permissions, review or baseline controls. Read access follows the user. Write actions require authorisation. Material changes can be prepared on a branch so reviewers can inspect the diff and connected evidence before merge.
04 MCP FAQs
How compatible AI clients use Arc records and controlled actions.
With the user’s Arc permissions, a compatible MCP client can search and retrieve programme records including requirements, systems, interfaces, risks, changes, verification activities, tests and evidence. Responses retain record identifiers and relationship context so the source can be inspected in Arc.
Yes. Authorised clients can create records, update fields, connect related items and propose controlled engineering changes. Arc applies the same workspace permissions and change controls used in the product rather than giving the client unrestricted access.
An agent can prepare work on a change branch and explain the affected context. Reviewers inspect the diff and decide whether to approve and merge it. Arc keeps the accepted baseline under human control.
Yes. Clients can navigate upstream intent and downstream allocations, interfaces, implementation records, tests and evidence. They can also ask Arc to identify connected records that may need review when a requirement changes.
No. MCP gives compatible AI tools a controlled way to use Arc data and actions. Arc remains the place to inspect the complete engineering graph, manage access, review proposed changes and control programme baselines.
05 Use Arc through MCP
Try Arc directly, then connect it to a compatible MCP client for governed search, traceability and change work.
I will drop you an email shortly from luc@archelps.com to get your free trial set-up. Please check your spam in case my email lands there.