MCP Config

Opsera MCP Configuration Guide

Make sure you have:

• An active Opsera account with login credentials

• Cursor IDE or Claude Desktop installed

1

Step 1: Generate Your API Token

This token is your secure key that connects Opsera to your AI tool.

1. Log in to your Opsera portal

2. Click the 3 dots next to your profile name at the bottom-left corner.

3. Select Profile.

4. Click the Access Tokens tab

5. Fill in the token form:

   • Name: Give it a clear name like MCP Integration

   • Expires After: Choose your preferred duration (1 Month, 3 Months, 6 Months, or 1 Year)

   • Scope: Select API Access

6. Click Create

⚠️ Important: Copy your token immediately after creation. It will only be shown once. Store it somewhere secure like a password manager — you cannot retrieve it again.

Your token will look like this: eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9...

2

Step 2: Add Opsera to Your Tool

hoose the tool you're connecting to:

Option A — Cursor IDE

Find your config file:

• macOS / Linux: ~/.cursor/mcp.json

• Windows: %USERPROFILE%\.cursor\mcp.json

Open the file and paste this configuration:

Replace YOUR_API_TOKEN_HERE with the token you copied in Step 1. Save the file.

{
  "mcpServers": {
    "opsera-ai-agent": {
      "url": "https://agent.opsera.ai/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_TOKEN_HERE"
      }
    }
  }
}

3

Option B: Claude Desktop

1. Open Claude Desktop

2. Go to Settings → Developer → Edit Config

3. Add the same configuration block above

4. Save and fully restart Claude Desktop

5. Go back to Settings → Developer and confirm opsera-ai-agent shows as Connected

4

Step 3: Verify It's Working

Once configured, test the connection with a simple command in your tool:

"List available MCP tools."

You should see Opsera pipeline and tool operations appear in the response. If they do, you're connected and ready to go.

What You Can Do

Once connected, you can manage your entire DevOps lifecycle through plain language commands. The Opsera MCP server exposes the following capabilities:

MCP ToolWhat It DoesExample Command

Pipelines

opsera_pipelines

List, view, start, stop, clone, reset, and delete pipelines. Fetch run status and execution history.

"Get the current status of pipeline ID [id]."

create_pipeline

Create a new SDLC pipeline from a plain-language description, including steps, tools, agents, and quality gates.

"Create a 3-step pipeline called 'Payments Deploy' from GitHub repo payments-service on branch main."

Tasks

opsera_tasks

Manage standalone tasks — create, list, update, run, and monitor task execution outside of a full pipeline.

"List all active tasks in the workspace."

Tools & Integrations

opsera_tools

View and manage connected tools (GitHub, SonarQube, Slack, etc.) registered in your Opsera account.

"List all tools connected to the org."

discover_tools

Discover available integrations and tool types that can be added to your Opsera environment.

"What tools are available for code quality scanning?"

Projects & Groups

opsera_projects

Create and manage projects that organize pipelines, tasks, and resources under a shared context.

"Show all pipelines under the 'Platform Engineering' project."

opsera_groups

Manage user groups, assign members, and control group-level access to pipelines and resources.

"List all members of the DevOps group."

Users

opsera_users

View and manage users in your Opsera organization — look up accounts, roles, and access permissions.

"Who are the admins in the org?"

Insights, Reports & Logs

opsera_insights

Retrieve analytics and trend data — pipeline success rates, build durations, deployment frequency, and more.

"Show me pipeline failure trends over the last 30 days."

opsera_reports_and_logs

Fetch detailed run logs, audit trails, and exportable reports for pipelines and tasks.

"Fetch the full logs from the last run of pipeline ID [id]."

Search & Settings

search_opsera

Search across all Opsera resources — pipelines, tasks, tools, users, and projects — using natural language.

"Find all pipelines that use SonarQube with a Grade B quality gate."

opsera_search_data_settings

Configure and manage search index settings and data preferences for your Opsera environment.

"Update the default search settings for the org."

Mappings

opsera_mappings

Manage environment and variable mappings used across pipelines, such as secret references, environment configs, and parameter bindings.

"List all variable mappings for the Production environment."

Sample:

API Reference

For teams integrating programmatically, here are the underlying endpoints the MCP tools call.

Learn more on each of the APIs here.

ActionEndpoint

Get registered tools GET /api/v2/registry/tools

Get step types GET /api/v2/tool/identifiers/module/pipeline-steps

Get step metadata GET /api/v2/step/metadata?identifier={stepType}

Browse repositories GET /api/v2/scm/repositories?gitToolId={id}

Browse branches GET /api/v2/scm/branches?gitToolId={id}&repoId={repo}

Validate pipeline POST /api/v2/pipeline/validate

Prompting Strategy

The 4-Line Prompt Formula

Line 1 — Identity & Source

Who is this pipeline, and where does it come from?

Cover the pipeline name, type, owning organization, Git repository, branch, and what triggers it.

Pipeline Name - Identifies what to create

Type (SDLC) - Defines pipeline category

Organization - Sets ownership context

GitHub Repo + Branch - Defines source code location

Push Event - Sets trigger behavior

Line 2 — Architecture & Order

How many steps, which tools, and in what sequence?

Define the number of steps, what tool each step uses, runtime versions, agent type, and how steps depend on each other.

Number of Steps - Defines pipeline length and scope

Shell Script / Command-Line - Defines the tool used per step

.NET Version Chain- Defines runtime environment per step

Linux / Windows Agent - Defines execution environment

Dependency Chain (→)- Enforces correct execution order

Line 3 — Configuration & Alerts

What are the quality gates, outputs, and notification rules?

Cover SonarQube settings, artifact outputs, notification channels, and any steps that behave differently or should be toggled inactive.

SonarQube + Grade - Defines code quality threshold

archive.zip Output - Defines the build artifact produced

Slack Notifications - Defines where alerts are sent

Active / Inactive Flags - Controls which steps actually run

Step Exceptions - Captures steps that behave differently

Line 4 — Output & Format

What should the response look like?

Specify the expected output format, schema requirements, and fields that must be present for the configuration to be complete and usable.

Valid JSON Output - Ensures the response is machine-readable

Standard Schema - Ensures structural accuracy

Step IDs + Dependencies - Ensures correct step mapping

Auto-Scale + Workspace Delete - Ensures config completeness

Example Prompts

Basic — Create a Simple Pipeline

Create an SDLC pipeline called "Hello World Deploy" for org Acme Corp, 
sourced from GitHub repo hello-world on branch main, triggered on push events.

1 step using Shell Script tool running on a Linux agent.

No quality gates. No notifications. Step active by default.

Return valid JSON following the standard Opsera pipeline schema 
with step IDs and dependency mappings included.

Intermediate — Multi-Step Build Pipeline

Create an SDLC pipeline called "Payment Service Deploy" for org Acme Corp, 
sourced from GitHub repo payment-service on branch main, triggered on push events.

6 sequential steps using Shell Script and Command-Line tools with .NET versions 
6 → 7 → 8 across steps, running on Linux agents, each step dependent on the previous.

SonarQube scan with minimum Grade D quality gate, output artifact as archive.zip, 
send Slack notifications on failure, Step 6 inactive by default with no artifact output.

Return valid JSON following the standard Opsera pipeline schema, including step IDs, 
dependency mappings, auto-scale enabled, and workspace delete set to true.

Advanced — Full Enterprise Pipeline

Create an SDLC pipeline called "Enterprise Core Release" for org Sample, 
sourced from GitHub repo core-platform on branch release/v3.2, 
triggered on push and pull request merge events.

8 sequential steps: Steps 1–3 on Linux agents using Shell Script with .NET 8, 
Steps 4–6 on Windows agents using Command-Line with .NET 7, 
Steps 7–8 on Linux agents for post-deploy validation. 
Each step dependent on the previous. Step 5 runs parallel to Step 4.

SonarQube scan on Step 3 with minimum Grade B quality gate. 
Output artifact as release-package.zip on Step 6. 
Slack notifications on any failure and on final success. 
Step 8 inactive by default. Steps 1 and 2 skip on weekends.

Return valid JSON following the standard Opsera pipeline schema, 
including all step IDs, dependency mappings, parallel execution flags, 
auto-scale enabled, workspace delete set to true, 
and environment variable placeholders for secrets.

Pipeline Management Commands

Beyond creation, use these prompt patterns for day-to-day pipeline operations:

# Run a pipeline
"Start pipeline ID [pipeline-id] and confirm execution status."

# Stop a pipeline
"Stop the currently running pipeline and retain the logs."

# Check status
"Get the current status and last run details for pipeline ID [pipeline-id]."

# Fetch logs
"Fetch the full logs from the last run of pipeline ID [pipeline-id]."

# Clone a pipeline
"Clone pipeline ID [pipeline-id] and name the copy [new-name]."

# Reset a pipeline
"Reset pipeline ID [pipeline-id] to baseline and clear all run history."

# Delete a pipeline
"Permanently delete pipeline ID [pipeline-id] after confirming it has no active runs."

Troubleshooting

Token error (401 Unauthorized) Your token may be incorrect, expired, or have the wrong scope. Go back to the Access Tokens page, generate a new one with API Access scope, update your config file, and restart your tool.

Opsera tools not appearing Make sure you fully closed and reopened your IDE or Claude Desktop — a partial restart won't pick up the new config. Also double-check that your mcp.json file is valid JSON with no missing brackets or commas.

Can't connect to Opsera Verify the URL https://agent.opsera.io is accessible from your network. If you're on a corporate VPN or firewall-restricted environment, you may need to whitelist this domain.

Token expired Return to your Opsera profile, generate a fresh token, update the mcp.json file with the new value, and restart your tool.