Cloud MCP file continuation reference

Read file chunks with the correct offsets and saved-file versions.

Customer Cloud MCP returns saved files in chunks.

Continuation fields

Field	Meaning
`encoding`	`utf-8` by default, or explicit `base64`.
`content`	This chunk's text or base64-encoded bytes.
`offset`	This chunk's starting position.
`nextOffset`	The next request's `offset`; `null` marks completion.
`version`	SHA-256 of the complete saved file's bytes.
`byteLength`	Complete saved-file size in bytes.
`manifestVersion`	Saved-file manifest version, including the workspace generation.

Encoding and offsets

Encoding	Chunk and continuation units
`utf-8`	Up to 16384 UTF-16 code units per chunk. Offsets count UTF-16 code units. Join text chunks in order.
`base64`	Up to 16384 original bytes per chunk. Offsets count decoded bytes. Decode each chunk, then join bytes in order.

Start at offset: 0; keep the same project, path and encoding. Follow nextOffset until null. Compare version and manifestVersion with the first response. If either changes, follow File changed during a read.

Saved-file sizes

File and encoding	Accepted complete-file size
Ordinary files; all `utf-8` reads	Up to 1 MiB.
`base64`: `alcad/documents/fab-scene/<name>.alcad-fab-scene.json`	Up to 4 MiB.
`base64`: `.gds` and `.gdsii` files	Up to 4 MiB.
`base64`: `rtl/schematic/results/<sha256>.json`	Up to 2 MiB.
`base64`: `evolve/analysis/results/<sha256>.json`, `alcad/circuit/rlc/results/<sha256>.json`, `alcad/circuit/digital/results/<sha256>.json`	Up to 8 MiB.

Use the paths returned by altifigence.workspace.files.list. If the tool says that the file requires binary encoding, start the file read again at offset: 0 with encoding: "base64".

See Connecting to MCP for setup.