Custom Auggie command to index and setup Context Connectors MCP

setup-context-connetors.md

description	argument-hint
Index GitHub repos with Augment Context Connectors and generate MCP server config	org/repo1#main, org/repo2#develop, org/repo3#v1.0

You are setting up Augment Context Connectors to index multiple GitHub repositories and configure an MCP server for cross-repo semantic search.

Step 1: Gather User Input

If $ARGUMENTS is empty or not provided, ask the user for:

1. A comma-separated list of repos in the format org_name/repo_name#ref_name (e.g., myorg/service-auth#main, myorg/service-payments#develop)
2. A GitHub Personal Access Token (PAT) with repo scope

If $ARGUMENTS is provided, use it as the repo list and ask only for the GitHub PAT.

Step 2: Parse the Repo List

Parse each entry from the comma-separated list into a tuple: (org_name, repo_name, ref_name).

For example, given: myorg/service-auth#main, myorg/service-payments#develop

Parse into:

• org_name=myorg, repo_name=service-auth, ref_name=main
• org_name=myorg, repo_name=service-payments, ref_name=develop

Derive the index name for each repo by joining org and repo with underscore: org_name_repo_name (e.g., myorg_service-auth).

Step 3: Index Each Repository

For each parsed repo, run the following command:

export GITHUB_TOKEN='<github-pat>' && npx @augmentcode/context-connectors index github \
  --owner <org_name> \
  --repo <repo_name> \
  --ref <ref_name> \
  --index <org_name>_<repo_name> \
  --store filesystem

Run these commands sequentially (one at a time), and report success/failure for each repo before moving to the next.

Step 4: Extract Augment Token

Run this command and capture its output:

auggie token print

The output will contain a line like:

SESSION={"accessToken":"...","tenantURL":"https://...","scopes":["read","write"]}

Parse the JSON from the SESSION= line to extract:

• accessToken — the Augment API token
• tenantURL — the Augment API URL

Step 5: Display MCP Configuration

Using all the collected values, display the following MCP server configuration snippet for the user to add to their ~/.augment/settings.json file:

{
  "mcpServers": {
    "Augment Context Connectors": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "@augmentcode/context-connectors",
        "mcp",
        "stdio",
        "--index <index_name_1> <index_name_2> <index_name_n>"
      ],
      "env": {
        "AUGMENT_API_TOKEN": "<accessToken>",
        "AUGMENT_API_URL": "<tenantURL>",
        "GITHUB_TOKEN": "<github-pat>"
      }
    }
  }
}

NOTES:

• the --index argument receives a list of index names separated by space (NOT commas). It is a one single param followed by the indexes names.
• <index_name_1> <index_name_2> <index_name_n> corresponds to the list of indexed repos (using the org_name_repo_name pattern)
• <accessToken> and <tenantURL> come from Step 4
• <github-pat> is the GitHub PAT provided in Step 1

Step 6: Summary

After displaying the config, provide a summary:

1. List all repos that were successfully indexed
2. List any repos that failed to index (if any)
3. Remind the user to:
   • Copy the JSON snippet into ~/.augment/settings.json (merge with existing config if needed)
   • Restart Auggie CLI or their IDE to pick up the new MCP server
   • The context-connectors MCP server will now allow semantic search across all indexed repos without naming them explicitly