vis-agent mcp serve: Vis as an MCP server on stdio, like claude mcp serve.
Each line on the input is one JSON-RPC 2.0 message, and each answer is one line
on the output. The caller gives the tools, which are the tools that a Vis agent
has. A tool call runs in one Vis session for the
workspace of :cwd. The first call creates the session, and later calls use it
again, so the variables of a block stay for the next block. The gateway runs each
block (POST /v1/sessions/:sid/run), and no model is called.
The output carries only protocol messages. Logs go to stderr.
`vis-agent mcp serve`: Vis as an MCP server on stdio, like `claude mcp serve`. Each line on the input is one JSON-RPC 2.0 message, and each answer is one line on the output. The caller gives the tools, which are the tools that a Vis agent has. A tool call runs in one Vis session for the workspace of `:cwd`. The first call creates the session, and later calls use it again, so the variables of a block stay for the next block. The gateway runs each block (`POST /v1/sessions/:sid/run`), and no model is called. The output carries only protocol messages. Logs go to stderr.
(serve! {:keys [in out err cwd tools debug?]})Serve MCP on in and out until the input ends, then answer exit code 0.
cwd is the workspace of the session, and tools are the tools that
tools/list answers. With debug?, each message goes to err, by default
stderr. Output of other code goes to *err*, so it cannot break the protocol.
Serve MCP on `in` and `out` until the input ends, then answer exit code 0. `cwd` is the workspace of the session, and `tools` are the tools that `tools/list` answers. With `debug?`, each message goes to `err`, by default stderr. Output of other code goes to `*err*`, so it cannot break the protocol.
cljdoc builds & hosts documentation for Clojure/Script libraries
| Ctrl+k | Jump to recent docs |
| ← | Move to previous article |
| → | Move to next article |
| Ctrl+/ | Jump to the search field |