Resources and HTTP routes

Every job’s files are also MCP resources, for clients that read resources rather than calling tools.

URI	What it returns
`jobs://index`	the jobs on this server, and the resource patterns below (for clients that list only static resources)
`job://{job_id}/status`	the full snapshot, the same as `job_status`
`job://{job_id}/events`	the event log, one JSON object per line
`job://{job_id}/log/{step}`	the end of a step’s log, such as `log/simpleFoam`
`job://{job_id}/postProcessing/{fo}/{file}`	a function object’s file from its latest time directory, capped at 4 MB
`job://{job_id}/figures/{set}/{file}`	a rendered image, such as `figures/p1/A.png`

HTTP route

A server running over HTTP also serves job files directly:

GET /jobs/{job_id}/files/{rel_path}

rel_path is relative to the job directory, so case/postProcessing/coeffs/0/coefficient.dat or figures/strip_A.png. Paths that escape the job directory are refused. Use it for anything too large to pass through a client’s context. On a server with OAuth, such as the hosted one, it needs the same bearer token as the MCP endpoint.