Questions tagged mcp
The MCP server, its tools, and wiring it into an editor.
20 questions
Example threads. These questions and answers were written by the SDODS team to document real problems and how to fix them. The people named are illustrative, not real users. Questions asked on this site are marked new.
Audit wants proof the agents cannot write to our repo — what do I actually show them?
Internal audit has heard the words "AI writes our tests" and wants a control, not a paragraph. Saying that agents only write proposals is exactly the kind of claim they will ask…
asked by Zeynep Arslansdods://screenshot resource returns No such screenshot for a run I can open in the UI
Asking the model to compare the before and after images of a failed step. It resolves the run fine — run_get_scenario comes back with the screenshot URIs — but reading one of…
asked by Zeynep Arslanheal_events, heal_locator_stats and sdods heal report — which do I use for what?
Building a small internal dashboard on top of the MCP server. There appear to be three ways to get at healing data and I cannot tell from the tool descriptions where the…
asked by Zeynep ArslanLocal model loops on agent generate, calling the same tool until it hits max turns
It calls feature_list, reads the result, calls feature_list again with the same arguments, and keeps going until the turn budget stops it. No proposal. It is not confused about…
asked by Paulo MendesEvery PR runs all 190 specs — is there a --since for sdods run?
PR checks take 40 minutes because we run the whole suite on every push, and a typical PR touches one module. I was expecting something like but that is not a flag. There is…
asked by Thom VasseurVS Code wrote .vscode/mcp.json but the chat still has no sdods tools
sdods mcp install vscode -p demo-shop reported that it created .vscode/mcp.json and the file looks right to me. Reloaded the window twice. The chat has the built-in tools and…
asked by Ingrid SolbergShould each tester run sdods mcp locally, or should we host one server for the team?
Eight testers, one repo, one staging environment. Right now everybody has stdio in their editor and their own checkout. It works, but nobody can ask about a run somebody else…
asked by Ingrid Solbergsdods mcp install claude dies with claude mcp add failed and writes nothing
On a locked-down work laptop: I do not have write access to that file and IT will not change it this quarter. Does the hint actually help me, or is it just a different way to fail?
asked by Helle BorgHow do I give the team coding agents read-only access to our SDODS server?
We host one SDODS server for the QA group. Everyone wants to point their editor at /mcp so they can ask about runs and coverage without a checkout. What I do not want is…
asked by Rasha HalabiMy mcp.servers block is ignored when the agent runs with the openai adapter
Project yaml has the usual: With --adapter claude the planner browses the app and the plan is full of real page names. With --adapter openai against our internal endpoint, the…
asked by Thom VasseurThe reference says 37 tools, the MCP guide says 33 to 35 — which number is right?
Writing an internal page about what our agents can reach and I cannot make the docs agree with themselves. The MCP tools reference is titled as the 37 tools. The MCP guide says…
asked by Koen VermeulenLocal model writes out the tool call as text instead of actually calling it
Running the generator on my own machine because the source cannot leave it: It does not call anything. It writes a paragraph about what it is going to do, then a line that looks…
asked by Tanvi BhattThe healer proposed a CSS locator when there is a perfectly good label — reject it?
Scenario broke, ran the healer, got a proposal. The patch swaps a page-object locator for a class chain: The input has a visible label reading "Promo code". So there is an…
asked by Anouk de VriesDoes sdods agent generate --dry-run call the model at all?
Before I point this at a paid key I would like to see what it is going to send. Does --dry-run do a cheap call, or no call?
asked by Bruno TeixeiraWhat should I actually look at in a proposal before running sdods proposals accept?
First week of using the generator role. sdods proposals show <id> gives me a diff and a manifest and I am reading it the way I read any PR, which I suspect is the wrong instinct…
asked by Bea Lindqvistcaps core,agent gives Unknown capability "agent" and I cannot find the real list
Right, plural. Fine. But is there a page that lists these, or do I read them out of the error every time?
asked by Fiona McAllistersdods mcp --http exits immediately with HTTP mode needs a bearer token
Trying to reach the server from a container on the same box, so stdio is out. Exit code 2. I did not expect auth on localhost. Is there a way to turn it off for a dev box?
asked by Nikhil SaneCan an SDODS agent commit the feature it wrote, or does a person always have to accept?
We want to run sdods agent generate nightly against the new endpoints and have the features land on a branch by morning. Right now every run leaves something in proposals/ that…
asked by Chandra Pillaisdods mcp install windsurf says Unknown client — is Windsurf just not supported?
Half the team uses Windsurf. Tried the obvious thing: Do I have to hand-write the config, or is there a flag I am missing?
asked by Yusuf Demirsdods mcp is registered but Claude Code lists no tools from it
Registered the server the way the guide says: claude mcp list shows sdods. But when I ask it to list projects it says it has no such tool and starts reading files instead. No…
asked by Ade Oyinlola