Repository navigation
docs: list 26 supported agent harnesses (#1991) - #2109
Conversation
Motivation: xerj init writes configs for three clients and prints setup lines for three more, while comparable tools publish a table of 20 or more, and readers use that list to decide whether a tool fits their setup. The gap is documentation, not code. What changed: a Supported agents section in the README, one row per client with its config file, config key, a note and a tested mark, the non-JSON shapes as snippets, and the same 26-row list on the docs CLI page. Verification: xerj v1.0.0-rc.97 on Windows answered initialize and tools/list over stdio, and OpenCode 2.1.2 registered the binary and reported connected after opencode mcp add wrote the project config. The other 25 rows are transcribed from each client's own documentation and are marked untested. Aider, T3 Code and Grok Bot are listed as excluded, with the reason. Signed-off-by: kclamano <klamano23@gmail.com>
|
Thanks for this contribution — we'd like to merge it. Before we can, we need a signed Contributor License Agreement from: @AK-Lmn It is a one-time thing per contributor, not per pull request. Read CLA.md, then open a small pull request adding your GitHub username to Once it is merged, comment |
|
@cla-bot check |
|
Re-checking the CLA status for every commit author on this pull request… |
main serves eighteen tools now and xerj-org#2102 adds a nineteenth; the tested row is about the client registering the server, not the count.
|
Thank you for your pull request and welcome to our community. We could not parse the GitHub identity of the following contributors: Xerj Merger.
|
|
Xerj Merger made the fix: reviewed the supported-agents table and snippets. They're docs only (README plus landing/docs/cli.html, no workflow changes), and it's good that every untested row is marked as such. I pushed one fix, 88b66ea: the OpenCode tested note said |
Closes #1991
What this changes, and why
The README's "Connect an agent over MCP" section showed one
mcpServersblock, andxerj initwrites configs for three clients and prints setup lines for three more. Comparable tools publish a
table of 20 or more, and readers use that list to decide whether a tool fits their setup, so the gap
was documentation rather than code.
This adds a Supported agents section to the README with one row per client - config file, config
key, the gotcha that trips people up, and a
testedoruntestedmark - plus the ten non-JSONshapes as snippets (Codex and Grok CLI TOML, OpenCode and Kilo command arrays, Continue and Goose
YAML, Amp, VS Code/Copilot, Zed, OpenClaw, DeepSeek Harness). The docs CLI page gets the same 26-row
list under a new "Supported agents" heading, and points at the README for the snippets.
Three candidates from the issue are deliberately absent, with the reason stated in the section:
Aider has no MCP client support at all (Aider-AI/aider#3314 is still open), T3 Code hosts an MCP
endpoint that other agents connect to rather than consuming one, and Grok Bot has no documented
local stdio registration. Grok's CLI does, so it gets its own row.
Evidence
The xerj side of the matrix, on Windows against the published
v1.0.0-rc.97x86_64 zip:One row was run end to end, OpenCode v2.1.2 on Windows:
That run also corrected the shape the docs describe: OpenCode v2 wants
mcp.servers.xerjwithtype: "local", onecommandarray, andenvironment(notenv), and it only reads the projectfile inside a project root, which is what the row now says.
Checks
cargo fmt --all- not applicable, no Rust touchedcargo test- not applicable, no Rust toucheddocs/CONTRIBUTION_REVIEW.md- not applicable, not a non-trivial code changeNot run:
cargo build,cargo test,cargo fmtand the ES-YAML suite, because the diff istwo documentation files.
Provenance
xerj mcpv1.0.0-rc.97 answeredinitializeandtools/listover stdio on Windows;opencode mcp addwrote the project config andopencode mcp listreportedxerj connected(OpenCode v2.1.2). The README table renders 26 datarows and the docs page list has 26 lines.
untested. Their paths andkeys come from each client's own documentation as of 2026-10-22, and no client was run. Trying any
row and finding the file absent or the key ignored falsifies that row. Two have less direct support
than the rest: Cursor's page was read through search-indexed copies because the site blocked direct
fetch here, and Claude Desktop's per-OS paths come from the MCP project docs and VS Code's
discovery table because Anthropic's current page no longer prints them.