DevTools Docs - MCP Servers & Agent Integrations

DevTools Documentation

MCP Servers & Agent Integrations

Cloud logging, MCP integrations, frontend tooling, and SSH workflows for production DevTools teams.

Account Settings Resources MCP Servers

SDK Repo

MCP Access

Cloud Logger

Frontend Tools

SSH Ready

MCP Servers & Agent Integrations

DevTools Included

DevTools subscribers can access two MCP servers: unified alshival_mcp and github-mcp. You can attach them directly to your own agents and tool-calling pipelines, and the built-in front-end Alshival assistant uses the same MCP tool surface.

No separate private tool stack exists for web chat. The same MCP contract powers your own agents and the in-product assistant.

Server Endpoints

Server	Endpoint	Primary Use
`alshival-mcp`	`https://mcp.alshival.ai/mcp/`	DevTools tools, account/resource operations, logging-related workflows.
`github-mcp`	`https://mcp.alshival.ai/github/`	GitHub tool access through delegated user context.

Authentication Requirements

X-API-Key: required for both MCP servers.
X-User-Username (recommended) or X-User-Email: required for user-scoped access.
GitHub MCP requires your DevTools account to have a connected GitHub login.

How Alshival Uses MCP in the Front End

In DevTools web chat, Alshival uses these same MCP servers with your user-scoped context. That is how the assistant can safely act on your resources, your sharing graph, and your DevTools social profile.

What you ask in chat	MCP tools Alshival uses	Outcome
"Check why my API is failing"	`resource_list`, `resource_get`, `resource_logs`	Reads cloud logs from your resource DB and summarizes likely root cause.
"Share this resource with @alex as developer"	`resource_share`	Grants/revokes/updates role-based access for collaborators.
"Find DevTools users and collaborate"	`search_users`, `social_interaction`	Discovers users, follows/unfollows, and manages collaborator mode.
"Set log alerts for this resource"	`resource_notification_settings`	Reads/updates APP/SMS/EMAIL notification settings per resource.

Subscriber MCP Tool Catalog (Core)

These are the primary tools exposed for DevTools subscriber workflows.

Area	Tools	What they enable
Knowledge + Directory	`search_kb`, `search_users`	Query platform/user knowledge and discover DevTools users.
Resource Lifecycle	`resource_list`, `resource_get`, `resource_upsert`	List, inspect, create, and update resources.
Cloud Logs	`resource_logs`, `resource_log_ingest`	Read and write structured cloud logs for resources.
Resource Sharing	`resource_share`	Grant, revoke, and role-manage collaborators.
Notifications	`resource_notification_settings`	Configure per-resource APP/SMS/EMAIL alert preferences.
Social Graph	`social_interaction`	Follow/unfollow users and activate collaborator mode.
Account	`get_account_settings`, `update_account_settings`	Read/update profile, model preference, and notification settings.

OpenAI Responses Tool Specs (Python / Raw JSON)

import alshival
from openai import OpenAI

client = OpenAI(api_key='your_openai_key')

tools = [
    alshival.mcp,
    alshival.mcp.github,
]

response = client.responses.create(
    model='gpt-5.2',
    input='Check my open PRs and summarize deployment risk.',
    tools=tools,
)

Requires OPENAI_API_KEY, ALSHIVAL_API_KEY, and ALSHIVAL_USERNAME (or ALSHIVAL_EMAIL) as environment variables.

{
  "type": "mcp",
  "server_label": "alshival-mcp",
  "server_url": "https://mcp.alshival.ai/mcp/",
  "require_approval": "never",
  "headers": {
    "x-api-key": "your_devtools_api_key",
    "x-user-username": "your_username"
  }
}

{
  "type": "mcp",
  "server_label": "github-mcp",
  "server_url": "https://mcp.alshival.ai/github/",
  "require_approval": "never",
  "headers": {
    "x-api-key": "your_devtools_api_key",
    "x-user-username": "your_username"
  }
}

Manual MCP Check with curl

curl -sS https://mcp.alshival.ai/mcp/ \
  -H "Content-Type: application/json" \
  -H "X-API-Key: $ALSHIVAL_API_KEY" \
  -H "X-User-Username: $ALSHIVAL_USERNAME" \
  -d '{"jsonrpc":"2.0","id":"1","method":"tools/list","params":{}}'

Using MCP with Other Agents

You are not limited to the built-in Alshival agent. Any agent framework that supports OpenAI Responses-style MCP tools can use these endpoints. Keep headers user-scoped so capability gating and role checks are applied correctly.

In practice, this means your custom agent can replicate the same front-end workflows: inspect logs with resource_logs, manage sharing with resource_share, and coordinate with users through search_users + social_interaction.