Other formats agents might prefer:
markdownjsonllms.txt

Agent? You probably want markdown or json, or Pod over MCP.

Reported issues for apify-mcp-server

Pod holds 24 of 147 problems reported by people outside the maintainer team. Issues filed by the project's own owners, members and collaborators are excluded entirely — a maintainer's release checklist is not a warning to a prospective user.

Back to apify-mcp-server.

Most discussed

fix: server card and server.json claim authentication is always required

ℹ️ Maintainers note: there was already discussion about changing the server.json authorization metadata to always required in https://github.com/apify/apify-mcp-server/pull/1174. The main issues is that the server.json spec does not support setting that authentication is required under some condition and we advertise in the server.json only the main mcp.apify.com URL that currently requires token. The MCP server might not require auth when the ?tools= is specified with tools that do not

Read the thread · 2026-07-28 · closed · external user · 5 comments

[Bug]: Unable to connect

Server type

Local (stdio via npx @apify/actors-mcp-server)

MCP Client

Claude Desktop

Operating System

None

What happened?

says can't connect api tried so many times

Steps to reproduce

cant connect api

Node.js version

No response

Apify MCP Server version

No response

MCP server configuration

apify

Error logs


Read the thread · 2026-03-29 · closed · external user · 5 comments

[Bug]: Install Issue

Server type

Local (stdio via npx @apify/actors-mcp-server)

MCP Client

Claude Desktop

Operating System

macOS

What happened?

Won't let me install for some reason.

M3 Pro 16 Gigs Ram Tahoe 26.3.1

Under requirements: <!>macOS, Windows, Linux (is red) and won't let me move forward.

Thank you for any tips.

Joey

Steps to reproduce

Open claude desktop settings, integrations, find apify macOS, Windows, Linux is in red, cannot install.

Node.js version

_No respon

Read the thread · 2026-03-23 · closed · external user · 6 comments

Unable to connect to extension server. Please try disabling and re-enabling the extension.

Trying to connect Apify to claude and i am receiving the below message after installing apify-mcp-server

Unable to connect to extension server. Please try disabling and re-enabling the extension.

Read the thread · 2026-03-14 · closed · external user · 11 comments

MCP call-actor tool should handle dataset previews better

Hello Apify developers!

This is an opinionated suggestion based on my tests of the Apify MCP server.

Context: I'm trying to let Claude Code use everything Apify has to offer using the new plugins and skills Current feelings: really nice scraper discovery, scraper info. Oauth with the remote server is super clean, love it as no secrets are needed anywhere! During the testing with CC my feeling was that the MCP mainly expects the users to use apify-slash-rag-web-browser and not so much the

Read the thread · 2025-10-18 · closed · external user · 8 comments

MCP server seems to only accept required fields when generating input to run actors

Edit: If it's actually a bug (and not a feature ;-) ) it was created by https://github.com/apify/actors-mcp-server/pull/158/commits/25ca1558e1b5ac2e4d39f57d93753b4341f00746 Edit 2: Happy to confirm that forcing the loading of the previous version 0.2.9 in my Claude config file is a functional workaround.

I use bebity/linkedin-jobs-scraper daily and here is a typical input: { "location": "United States", "proxy": { "useApifyProxy": true, "apifyProxyGroups": [

Read the thread · 2025-07-10 · closed · external user · 5 comments

May I know why we decide not to expose the other tools for more granular control of actors?

Curious to understand why we only expose few tools now. It makes me hard to check , review the run I have. Is this a resource concern? is it possible to bring them back before we have dataset in resources? the data response right now is very limited.

Read the thread · 2025-07-09 · closed · external user · 16 comments

Fix Actor search tool description

Context: https://apify.slack.com/archives/C08N1LWC596/p1745883305680539

Read the thread · 2025-04-29 · closed · outside contributor · 5 comments

Most recent

Connector description is empty, so capabilities are unknown before enabling

Problem

In the client's connector list, the Apify connector shows an empty description. Nothing states what it can do before it is enabled, so an agent recommends it on the name alone, the user enables it, and only then does the task turn out to be unsupported (deploying Actor source). The round trip is wasted on missing metadata alone.

Expected

The listing carries a description of what the connector covers (search and run Actors, read datasets and key-value stores, manage tasks) and wh

Read the thread · 2026-08-28 · open · external user · 0 comments

Map the 'apify-ai' MCP client name to APIFY_AI request-origin attribution

The Apify AI agent service (apify/apify-ai-agent, the Mastra service replacing Console AI chat's backend) connects to mcp.apify.com with MCP clientInfo.name = 'apify-ai'. getRequestOriginForClient() (src/utils/mcp_clients.ts) exact-matches APIFY_AI_CLIENT_NAME (apify-console-ai-chat), so the service's Actor runs are attributed as generic MCP instead of APIFY_AI in run.meta.origin and origin-filtered listings.

Ask: also map `apify-

Read the thread · 2026-08-26 · open · outside contributor · 0 comments

MCP spec conformance: 5 requirement(s) violated (via @hasmcp/mcp-spec-test) — spec 2025-11-25

When run over stdio with npx @apify/actors-mcp-server and no APIFY_TOKEN set, the server prints APIFY_TOKEN is required but not set... to stderr and exits immediately, before ever completing an MCP handshake. Testing specifically against the older 2025-11-25 spec revision (this server negotiates both 2026-07-28 and 2025-11-25), @hasmcp/mcp-spec-test finds 5 requirement violations — fewer than under 2026-07-28 because this revision's suite doesn't have a separate server/discover

Read the thread · 2026-08-24 · open · external user · 0 comments

MCP spec conformance: 10 requirement(s) violated (via @hasmcp/mcp-spec-test) — spec 2026-07-28

When run over stdio with npx @apify/actors-mcp-server and no APIFY_TOKEN set, the server prints APIFY_TOKEN is required but not set... to stderr and exits immediately, before ever completing an MCP handshake. Because of this, @hasmcp/mcp-spec-test's conformance suite (which starts the server as a plain stdio child process, with no credentials configured) can't get past initialize/server/discover at all, and 10 requirement checks fail as a result — mostly cascading consequences of "no

Read the thread · 2026-08-24 · open · external user · 0 comments

initialize handshake blocks on unbounded network call, causing client-side timeout in shared-pool sessions (Claude Desktop Cowork/Code)

Environment Extension: apify-mcp-server (Claude Desktop extension ant.dir.gh.apify.apify-mcp-server) Version: 0.14.3 OS: Windows 11 Node: bundled/"built-in Node.js" runtime used by the Claude Desktop extension host Server args: dist/stdio.js --tools actors,docs,apify/rag-web-browser Client: Claude Desktop, specifically the separate server copy it spawns for Cowork/Code sessions ("shared-pool") Summary The MCP initialize request handler performs a live network call to the Apify platform API (fetc

Read the thread · 2026-08-21 · open · external user · 1 comment

feat: Add task tools to the default eval tools

Problem or motivation

The eval harness only loads a fixed set of tool categories, and the task tools had to be added to that list by hand for the new cases to work. They should be part of the default tools the evals run against so this doesn't have to be remembered per category.

What would good look like?

No response

Alternatives considered

No response

Read the thread · 2026-08-19 · open · outside contributor · 1 comment

feat: add integration tests for Actor tasks

Problem or motivation

The task tools (get, create, update, publish, unpublish) shipped with unit tests only. There's no integration coverage in tests/integration/suite.ts, so nothing verifies they work against a real API. Requested during review of #1235

Read the thread · 2026-08-19 · open · outside contributor · 0 comments

[Bug]: search-actors treats an exhausted page as no matching Actors

Server type

Local (stdio via @apify/actors-mcp-server)

MCP Client

Other: a minimal MCP stdio contract client

Operating System

macOS

What happened?

search-actors returns the same no-results guidance for two different states:

  1. the search has no matching Actors; and
  2. the query has results, but the requested offset is past the last page.

For an exhausted page, the successful response says that no Actors matched and requires a broader query. It does not include the offs

Read the thread · 2026-08-18 · open · external user · 0 comments

fix: render tool descriptions conditionally on session tool set instead of prose hedges

Problem

Tool descriptions are static module-level strings built once at import (e.g. WIDGET_ADDENDUM in src/tools/actors/call_actor.ts), so a cross-tool reference to a tool that may be absent from the resolved session's tool set gets patched with a prose hedge ("...if that tool is available in this session") instead of actually being omitted. PR #1261 added/extended ~20 of these across 14 files. The hedge still names an absent tool and burns tokens on every session — including the comm

Read the thread · 2026-08-17 · closed · outside contributor · 0 comments

Feature Suggestion: Optional token metering & paid API key support via neuforge-pay

Hi @apify,

Love the work on apify/apify-mcp-server! As usage grows across AI agent frameworks, server compute costs can start adding up quickly. We've also seen developers report that their autonomous agents accidentally burn through thousands of dollars in a single session because traditional payment rails lack built-in agent guardrails.

Would you be open to adding an optional usage metering & billing decorator using neuforge-pay? It acts as a proactive Spend Firewall and Merchant of

Read the thread · 2026-08-16 · open · external user · 0 comments

fix(evals): Cases without context still get a "previous interaction: {}" turn

What happens

createOpenRouterTask() serializes the context before checking it:

// evals/evaluation_utils.ts:67
const context = JSON.stringify(example.input?.context ?? {});
...
// evals/evaluation_utils.ts:74
if (context) {

JSON.stringify({}) is "{}", so the guard is always true and every case gets an extra user turn before the query. No case in evals/test_cases.json sets context today, so this applies to all 92.

Captured from the real createOpenRouterTask() with th

Read the thread · 2026-08-15 · closed · external user · 1 comment

fix: Workflow eval case targets removed Actor apify/actors-mcp-server

Issue

The Actor apify/actors-mcp-server no longer exists:

The workflow eval case fetch-details-mcp-tools-then-call is built entirely around it. The case now always fails

Several places still mention this Actor as an example Actor and this is no longer valid

Fix

Adjust the case to point at an Actorized MCP server that exists. apify/example-mcp-server

Separately, apify/actors-mcp-server is still named as the example in tool descriptions the model reads (`src/tools/actors/call_act

Read the thread · 2026-08-07 · open · outside contributor · 0 comments

chore: Move evals to langfuse

Currently evals are run locally and the results are printed to terminal or to a file.

We would like the test cases to be datasets and for traces and results to be visible on langfuse.

Read the thread · 2026-08-06 · open · outside contributor · 0 comments

spec: Workflow evals: multi-dimension rubric + deterministic checks

Problem

evals/workflows/ judges every multi-turn agent conversation with a single binary PASS/FAIL, produced by an LLM judge that only ever sees tool calls + arguments + agent text

There's also no deterministic ground truth anywhere in this harness, everything is left to LLM judgment.

Proposed solution

Define a fixed 6-dimension rubric for the standalone judge, with each dimension independently scored PASS/FAIL + reason: **toolSelection, argumentCorrectness, resultUtilization, tas

Read the thread · 2026-08-05 · open · outside contributor · 1 comment

test: Add MCPJam apps and tasks conformance coverage

Problem

Two conformance holes exist that the official runner (@modelcontextprotocol/conformance with pnpm run test:conformance) cannot close.

1. The MCP Apps surface is effectively untested end to end. scripts/check_widgets.ts only asserts that a registered widget's built JS file exists.

2. The tasks extension cannot be reached by the official runner at all. Our tasks implementation therefore has unit coverage only.

MCPJam ships conformance runners for both surfaces. A ful

Read the thread · 2026-08-05 · open · outside contributor · 1 comment

Integration test suite rework: split suite.ts into flat per-concern cases

What

Split tests/integration/suite.ts (3292 lines, 115 cases, one file) into per-concern modules. Share the majority of cases with apify-mcp-server-internal instead of hand-duplicating ~40 of them there.

Why

One file, 115 cases, unmaintainable by more than one person at a time. Internal re-implements a third of the same assertions against a live deploy — drifts, doubles maintenance cost, no shared source of truth.

Structure — flat, not mirrored to src/

tests/integration/c

[Read the thread](https://github.com/apify/apify-mcp-server/issues/1199) · 2026-08-05 · closed · outside contributor · 1 comment

The remaining reports are on [the project's issue tracker](https://github.com/apify/apify-mcp-server/issues).