Enterprise analysis platforms are keyed on organisation, team, repository and project identifiers that live in a web console. Most integrations require them up front, which means the first thing a user does is go hunting. Here the tools walk the hierarchy from what the token can see, and then write the config file from what they found. create_repo insisting that list_scm_auths be called first is the same instinct applied to credentials: the repository cannot be connected without one, so the tool text makes the ordering explicit rather than letting the call fail. Tool descriptions and parameters are in Chinese.
A client for Tencent Cloud Code Analysis covering both halves: the discovery and setup path that gets a repository connected, and the scanning path that runs analysis and returns findings.
- Discovery walks down the hierarchy — organisations, then teams, then connected repositories, then the analysis projects and available schemes under them — so configuration can be found rather than asked for.
- setup_config writes the tca-mcp.ini file from what discovery found, which is the step that otherwise involves copying identifiers by hand.
- Repository onboarding is credential-aware: create_repo's own description requires calling list_scm_auths first, and create_scm_account and create_scm_ssh add username-password or SSH credentials when none exist.
- An analysis project is defined as a scheme plus a branch, and create_project builds one from those two.
- start_scan runs the analysis, job_list and job_detail track it, tca_issue_list returns the unresolved issues, and tca_issue_report returns the report link.
TCA_TOKEN and TCA_USER_NAME, plus a tca-mcp.ini config file — which setup_config can generate once discovery has found your organisation, team and repository.
One command plus a key — npx -y -p tca-mcp-server@latest tca-mcp-stdio, then supply credentials