BigPanda MCP server: Biggy tools and action plans

How AI clients sign in to Biggy, what its named tools cover, how the action plan endpoint differs, and how BigPanda consumes your servers in the other direction.

Checked 6 October 2026 · from BigPanda's Connect to Biggy docs, API spec and live endpoints

One of 91 guides to monitoring and developer tools. Each covers what an AI coding agent can get out of the tool through MCP, the protocol that lets the agent query the tool directly.

What is the BigPanda MCP server?

BigPanda correlates alerts into incidents. Biggy is its AI assistant, and an action plan is a configured piece of work Biggy knows how to do, such as troubleshooting an issue from historical data.

The Connect to Biggy server puts that assistant inside your own AI tool. ask_biggy is the full Biggy agent as one tool. search_knowledge is one of the narrow read tools BigPanda calls direct tools.

The API endpoint exposes action plans directly. An external agent lists them, picks one from its description, and sends it a question in English. The action plan does the work and answers.

Elsewhere in this directory, a tool is usually an API call with typed parameters. Here the main tool is another agent, and the parameter is a sentence.

BigPanda sits on both ends of MCP:

Each row is configured in a different place. A search result about connecting an MCP server to Biggy describes the client direction.

BigPanda's subject is the incident. Its reach depends on what Biggy and your enabled action plans can do.

In practice, BigPanda answers incident questions at whatever depth Biggy and its action plans support. Two organizations connecting the same server can see different tools, because roles and the admin policy scope them.

What can an AI agent do with BigPanda?

Two endpoints carry different tool lists. AI clients that sign in to mcp.biggy.io/mcp get ask_biggy, search_knowledge and get_conversation, scoped to the person's role. The API endpoint at api.biggy.io/mcp turns each enabled action plan into one tool. The main tools on both take a question in English, so an agent asks incident questions that Biggy and your plans already cover.

BigPanda MCP server limits: where answers come back incomplete

BigPanda's first limitation is unusual. What an agent sees depends on the endpoint, the person and the policy.

How to configure BigPanda MCP for agents

1. 1

   Pick the endpoint first

   People connecting their own AI tool use mcp.biggy.io/mcp and sign in. Automation built on action plans uses api.biggy.io/mcp with a user API key.

2. 2

   Check the admin policy before rollout

   AI clients are off by default. An administrator enables them on the AI clients page and chooses which tool families the server registers.

3. 3

   Keep actions behind approval

   ask_biggy_act appears only when an administrator allows actions and the connection carries the biggy:act scope. Each action then waits for a person to approve it.

4. 4

   On the API endpoint, audit the action plans

   They are the tool list there. Whatever is enabled is what an external agent can see and call.

5. 5

   Expect stateless HTTP on the API endpoint

   It handles initialize, tools/list and tools/call over POST. GET and DELETE return 405, because there are no sessions and no server-sent event streams.

6. 6

   Decide the reasoning effort deliberately

   Every action plan tool accepts low, medium or high, and the specification leaves the default unstated.

7. 7

   Write the whenToSelect field carefully

   It becomes part of the tool description an external agent reads when choosing an action plan.

8. 8

   Keep the client direction separate

   Adding an MCP server to Biggy is a different feature, configured in the web app, with its own trust policy.

Paste this into your agent’s instructions

The main tool here is an agent, whether it is ask_biggy or an action plan. Two layers of inference between the question and the alert data is worth stating.

When you answer using BigPanda MCP tools, state:
- Which tool or action plan you called, and the query text
  you sent it. The tool takes free text, so the wording
  matters.
- That the answer came from Biggy or an action plan rather
  than from a direct query against alert data.
- The reasoning effort you requested, if you set one.
When a tool returns little, report exactly that. State the
tool or action plan you called and what it returned.
Write your answer in ASD-STE100 Simplified Technical
English. Use short sentences with one idea in each.

Do you need the MCP server at all?

Not for anything that runs on a timer. The alert and incident endpoints of the BigPanda API take typed parameters, and they are the direct route for a report you already know how to shape.

For alert correlation without the action plan layer, PagerDuty and Opsgenie expose incident data as structured tools instead of as prose.

Where the question is about metrics rather than incidents, Datadog, Grafana and Prometheus answer it far more cheaply.

BigPanda also publishes an A2A interface for its Alpha agent, using JSON-RPC 2.0 for messages, streamed responses and task status. That is the route for agent-to-agent work rather than tool calls.

With MCP, work your team has already configured becomes callable from outside, without rebuilding any of it as an API integration.

Call tools/list first, as the person or key that will run the agent. The answer depends on your own role, policy and action plans.

BigPanda MCP server setup

The server is hosted, so there is nothing to install. Every AI tool connects to https://mcp.biggy.io/mcp, and BigPanda's Connect to Biggy page generates the snippet for each one.

Each block below adds the server by URL. On first use, the tool opens the Biggy sign-in, and you connect as yourself with your Biggy permissions.

Some tools show an Advanced: use a personal key instead toggle on that page. It switches the snippet to a personal key on the /mcp/key endpoint, for a client that cannot complete the sign-in.

The action plan endpoint at api.biggy.io/mcp is a separate route for automation. It takes a user API key, and the second block shows it.

claude mcp add --transport http bigpanda \
  https://mcp.biggy.io/mcp

curl -sS -X POST "https://api.biggy.io/mcp" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -H "Authorization: Bearer <api-key>" \
  -d '{"jsonrpc":"2.0","id":2,
       "method":"tools/list","params":{}}'

{"mcpServers":{"bigpanda":{"type":"http",
  "url":"https://mcp.biggy.io/mcp"}}}

[mcp_servers.bigpanda]
url = "https://mcp.biggy.io/mcp"

Every other client

Each block below is the configuration for one client, with the file path and the key that client expects.

{"mcpServers":{"bigpanda":{"command":"npx",
  "args":["-y","mcp-remote","https://mcp.biggy.io/mcp"]}}}

{"mcpServers":{"bigpanda":{
  "httpUrl":"https://mcp.biggy.io/mcp"}}}

{"servers":{"bigpanda":{"type":"http",
  "url":"https://mcp.biggy.io/mcp"}}}

{"mcpServers":{"bigpanda":{
  "serverUrl":"https://mcp.biggy.io/mcp"}}}

{"context_servers":{"bigpanda":{
  "url":"https://mcp.biggy.io/mcp"}}}

{"mcpServers":{"bigpanda":{"type":"http",
  "url":"https://mcp.biggy.io/mcp"}}}

{"amp.mcpServers":{"bigpanda":{
  "url":"https://mcp.biggy.io/mcp"}}}

{"mcpServers":{"bigpanda":{"type":"streamableHttp",
  "url":"https://mcp.biggy.io/mcp"}}}

extensions:
  bigpanda:
    type: streamable_http
    uri: https://mcp.biggy.io/mcp

{"mcpServers":{"bigpanda":{"command":"npx",
  "args":["-y","mcp-remote","https://mcp.biggy.io/mcp"]}}}

{"mcpServers":{"bigpanda":{
  "url":"https://mcp.biggy.io/mcp"}}}

{"mcpServers":{"bigpanda":{"type":"http",
  "url":"https://mcp.biggy.io/mcp"}}}

{"mcpServers":{"bigpanda":{
  "type":"streamable-http",
  "url":"https://mcp.biggy.io/mcp"}}}

name: BigPanda
version: 0.0.1
schema: v1
mcpServers:
  - name: bigpanda
    type: streamable-http
    url: https://mcp.biggy.io/mcp

{"mcpServers":{"bigpanda":{
  "url":"https://mcp.biggy.io/mcp"}}}

{"mcpServers":{"bigpanda":{"type":"http",
  "url":"https://mcp.biggy.io/mcp"}}}

claude mcp add --transport http bigpanda \
  https://mcp.biggy.io/mcp/key \
  --header "Authorization: Bearer <personal-key>"

Integrations  ->  + Add MCP
  Service Type:   Observability | ITSM | Context Retrieval
  MCP Server URL: https://your-server.example.com/mcp
  Bearer Token:   <token Biggy uses to call your server>
  Deployment:     Cloud | On Prem
  Trust Policy:   Allow only selected tools (recommended)

Read-only access, permissions and security

• Scope comes from the person, the policy and the plans

  On mcp.biggy.io, each connection acts as the person who signed in, within their Biggy role, and every call is recorded under that account. On api.biggy.io, every enabled action plan becomes callable.

• Free text is the main parameter

  ask_biggy and every action plan tool take a sentence, so the agent on the other side decides what to do with it.

• Actions stay behind approval

  ask_biggy_act needs the biggy:act scope and an administrator setting. Each action pauses until a person approves it. Revoking a connection under Your connections signs that tool out immediately.

• In the client direction, prefer the allowlist

  BigPanda offers two trust policies. Allow only selected tools is the one it marks as recommended, and new tools then need explicit approval. Auto-trust all tools allows every tool the server exposes, including ones added later.

• The on-prem option crosses your firewall deliberately

  Marking an integration as on-prem makes it eligible for the Biggy Relay Client, which is how it reaches internal infrastructure.

• The API endpoint takes a user API key

  The specification states that this API uses the User API Key type of token, so the credential reaches beyond MCP. The Connect to Biggy server uses a per-person OAuth sign-in.

Troubleshooting

`401` with `missing_bearer`

The client has not signed in yet. Run its sign-in step, such as /mcp in Claude Code, or use a personal key on the /mcp/key endpoint.

Tools are refused after sign-in

AI clients are off by default. An administrator turns them on under Manage, Integrations, AI Clients, with the Allow AI clients setting.

`405 Method Not Allowed`

The API endpoint accepts POST only. The specification states that GET and DELETE return 405, because stateless mode uses neither sessions nor server-sent event streams.

`tools/list` returns nothing useful

On the API endpoint, the list is your enabled action plans. If it is short or empty, check the BigPanda action plan configuration. On mcp.biggy.io, check the tool families on the AI clients page.

The agent picks the wrong action plan

Tool descriptions are generated from each plan's description and whenToSelect fields. Rewriting those fields is the supported way to change how an external agent chooses.

Answers are shallow or too slow

Every action plan tool accepts a reasoningEffort of low, medium or high. It is optional, so a client that omits it takes whatever the server defaults to.

The response is hard to parse

Set responseFormat to plain-text. The default documented in the schema table is markdown.

Biggy cannot reach your internal server

Set the deployment type to On Prem. That marks the integration as eligible to connect through the Biggy Relay Client, which is the documented route to infrastructure behind a firewall.

BigPanda MCP server: Reference

What the vendor documents

The facts below come from BigPanda's Connect to Biggy and AI clients pages, last modified 22 September 2026, and from its API specification and MCP integration documentation. Public discussion of the endpoints is thin.

“Once connected, your AI tool sees the same set of tools named ask_biggy, search_knowledge, and get_conversation, scoped to your role and what your organization has exposed.”

“Each enabled action plan in your organization becomes an MCP tool. Tool descriptions are derived from the action plan's description and whenToSelect fields, giving external agents enough context to choose the right tool autonomously.”

“Allow only selected tools (Recommended): Pick exactly which tools Biggy can call. New tools require explicit approval.”

“GET and DELETE calls return 405 Method Not Allowed (stateless mode does not use sessions or server-sent event (SSE) streams).”

BigPanda's Connect to Biggy and AI clients pages and its API specification were read on 6 October 2026, and both endpoints were probed on the same date. The action plan tool count is per organization, so none is claimed for that endpoint.

Should you connect an agent to BigPanda?

Yes. Best for incident questions phrased in English, from people whose Biggy roles you have reviewed.

Strongest fit

BigPanda customers who want Biggy inside the AI tools their engineers already use.

Main advantage

Each person signs in as themselves, and actions wait for approval. Existing action plans are also callable through the API.

Main weakness

Most of the depth sits behind one free-text tool, ask_biggy, and the action plan list differs in every organization.

Operational risk

Free text is the main input, so Biggy decides what a question means before it acts on it.

Frequently asked questions