pod
Other formats agents might prefer:
markdownjsonllms.txt

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

Reported issues for Paper Chaser MCP

Pod holds 17 of 23 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 Paper Chaser MCP.

Most discussed

[aw] Test Paper Chaser MCP failed

Workflow Failure

Workflow: Test Paper Chaser MCP
Branch: master
Run: https://github.com/joshuasundance-swca/paper-chaser-mcp/actions/runs/24641694952

⚠️ Engine Failure: The copilot engine terminated before producing output.

Error details:

  • Authentication failed (Request ID: B441:8CEEC:28CF06:362836:69E564BD)

💡 This repository is a fork. If this failure is due to missing API keys or tokens, note that secrets from the parent repository are not inherite

Read the thread · 2026-04-19 · open · external user · 1 comment

[agentic-test] Provider-specific CORE search flattens upstream failure into a no-results style payload

Smoke-mode MCP UX review found a concrete defect in the provider-specific CORE path.

Reproduction

  1. Call search_papers_core(query="transformer architecture", limit=3).
  2. Observe the response returns data: [] and brokerMetadata.providerUsed: "none".
  3. Observe brokerMetadata.attemptedProviders[0].status is "failed" with a raw CORE 500 Internal Server Error message.
  4. Observe brokerMetadata.nextStepHint says "No provider returned results. Try broadening the query...".

E

Read the thread · 2026-03-18 · closed · external user · 1 comment

[agentic-test] search_papers_match misses canonical known-item lookup for Attention Is All You Need

Summary

search_papers_match(query="Attention Is All You Need") returned a structured no-match payload during the agentic MCP review, even though the same MCP surface could resolve the paper immediately through other supported paths.

This is a high-impact UX regression because known-item lookup is supposed to be the low-friction path for a famous paper title. It also forces unnecessary agent round trips: a task that should take 1 call now takes 2-3 calls plus fallback reasoning.

Reprodu

Read the thread · 2026-03-18 · closed · external user · 1 comment

[agentic-test] Normalize CORE-backed publicationTypes to a list in paper search responses

Summary

During the smoke-path MCP review, CORE-backed paper search results violated the promised normalized response shape by returning publicationTypes as a scalar string instead of a list.

This is high-impact for agent UX because agents reasonably infer a stable cross-provider shape from the tool names/descriptions and README. A type flip on a common field creates avoidable branching, weakens tool intuitiveness, and makes provider pivots less safe.

Reproduction

Tool calls used in thi

Read the thread · 2026-03-18 · closed · external user · 1 comment

[agentic-test] get_author_papers rejects documented open-ended publicationDateOrYear filter

Smoke review found a reproducible break in the documented author workflow.

Context:

  • Effective mode: smoke
  • Repository workflow run: 23214409718
  • Golden path/docs currently advertise search_authors -> get_author_info -> get_author_papers(author_id=..., publicationDateOrYear="2022-", limit=...)

Reproduction:

  1. Call search_authors(query="Yoshua Bengio", limit=3).
  2. Take the first result's authorId (1751762 in this run).
  3. Call get_author_info(author_id="1751762").
  4. Call `ge

Read the thread · 2026-03-17 · closed · external user · 1 comment

[agentic-test] search_papers_match is brittle to common title capitalization on canonical paper lookups

Summary

During the smoke UX review, search_papers_match failed for the common query "Attention Is All You Need" even though the paper is clearly available through the same MCP surface.

This breaks the known-item golden path in a way that feels surprising to a low-context agent: a famous exact-title lookup returned a structured no-match, while a nearly identical capitalization variant succeeded.

Why this matters

search_papers_match is supposed to be the low-friction entry point for

Read the thread · 2026-03-17 · closed · external user · 1 comment

[agentic-test] search_papers_bulk exposes both raw token and wrapped nextCursor, making pagination contract ambiguous

Summary

During the smoke review, search_papers_bulk(query="graph neural networks", limit=5) returned both:

  • a top-level raw token
  • a structured pagination.nextCursor

The documented/desired workflow is to pass back pagination.nextCursor unchanged, but exposing the provider token alongside it makes the contract feel ambiguous to new agents.

Why this matters

This is a high-traffic path for exhaustive retrieval. The current response shape weakens an otherwise good pagination story

Read the thread · 2026-03-17 · closed · external user · 1 comment

[aw] No-Op Runs

This issue tracks all no-op runs from agentic workflows in this repository. Each workflow run that completes with a no-op message (indicating no action was needed) posts a comment here.

📘 What is a No-Op?

A no-op (no operation) occurs when an agentic workflow runs successfully but determines that no action is required. For example:

  • A security scanner that finds no issues
  • An update checker that finds nothing to update
  • A monitoring workflow that finds ev

Read the thread · 2026-03-16 · open · external user · 1 comment

Most recent

[agentic-test] Guided research false-grounds nonsense queries instead of abstaining

Summary

In a guided-profile review run (mode=comprehensive, requested tool_profile=guided, focus prompt regulatory (ecos, federal register, etc)), the no-results safety check failed: research(query="asdkfjhasdkjfh research paper nonsense", limit=3) returned a partially grounded result instead of abstaining.

The response treated the token nonsense as a valid semantic anchor and surfaced Neuro-Nonsense: Why Ulysses Contracts don't Compute in Brain-Computer Interface Research as gro

Read the thread · 2026-04-04 · closed · external user · 0 comments

[agentic-test] resolve_reference returns high-confidence bestMatch despite conflicting author/year/venue metadata

Summary

resolve_reference can return a status: resolved, resolutionConfidence: high, and a single bestMatch even when the same payload explicitly reports major conflicts on author, year, and venue. That makes the tool feel unsafe for low-context agents because the top-line contract says "usable match" while the details say "probably not this paper."

Reproduction

  1. Call resolve_reference(reference="Rockstrom et al planetary boundaries 2009 Nature 461 472") in the default guided

Read the thread · 2026-04-04 · closed · external user · 0 comments

[agentic-test] search_papers_bulk default ordering feels like a misleading continuation from search_papers

Smoke review summary:

  • Effective mode: smoke
  • Focus prompt: none supplied
  • Workflow run: 23248983861

Why this matters

search_papers currently encourages agents to continue with search_papers_bulk, but the default bulk first page for a normal topical query looks much less relevant than the discovery page. That makes the continuation feel misleading and increases agent round trips because the agent has to infer that bulk is not returning a relevance-ranked continuation.

Reprod

Read the thread · 2026-03-18 · closed · external user · 0 comments

[agentic-test] search_papers_match misses canonical known-item lookup for "Attention Is All You Need"

Smoke-mode MCP review found a high-impact golden-path regression in the known-item workflow.

Why this matters

search_papers_match is the primary low-round-trip tool for messy or exact title lookup. It failed on one of the most canonical papers in the corpus, which forces agents into undocumented fallbacks and makes the known-item path feel unreliable.

Reproduction

  1. Call search_papers_match(query="Attention Is All You Need").
  2. Observe the response:
    • matchFound: false

Read the thread · 2026-03-18 · closed · external user · 0 comments

[agentic-test] get_author_papers exposes a misleading fields parameter that rejects paper fields

Summary

During a smoke review of the MCP golden paths, get_author_papers worked functionally, but its argument contract was misleading enough to cause a first-try validation failure.

The tool returns paper records, so it is natural to request paper-shaped fields like title, year, and authors. Instead, the schema exposes a fields parameter inherited from AuthorInfoArgs, and validation rejects those paper fields with an error that only accepts author-profile fields.

Reproducti

Read the thread · 2026-03-18 · closed · external user · 0 comments

[agentic-test] search_papers_match misses exact title lookup for Attention Is All You Need

Summary

search_papers_match failed to resolve the exact-title query Attention Is All You Need, even though the paper is resolvable from the same MCP surface via get_paper_details(paper_id="arXiv:1706.03762") and is discoverable via search_papers(query="Attention Is All You Need").

This breaks the repo's known-item golden path and adds unnecessary agent round trips for a very common lookup.

Reproduction

  1. Call: search_papers_match(query="Attention Is All You Need")
  2. Observ

Read the thread · 2026-03-18 · closed · external user · 0 comments

[aw] Test Scholar Search MCP failed

Workflow Failure

Workflow: Test Scholar Search MCP
Branch: master
Run: https://github.com/joshuasundance-swca/scholar-search-mcp/actions/runs/23227785857

💡 This repository is a fork. If this failure is due to missing API keys or tokens, note that secrets from the parent repository are not inherited. Configure the required secrets directly in your fork's Settings → Secrets and variables → Actions.

Action Required

Option 1: Debug using any coding agent

U

Read the thread · 2026-03-18 · closed · external user · 0 comments

[agentic-test] search_papers no-results path can return irrelevant results without low-confidence signaling

Summary

During a smoke UX review, the brokered discovery path returned clearly irrelevant papers for a nonsense query instead of surfacing an empty or explicitly low-confidence result.

That is a meaningful agent UX defect: a low-context agent can reasonably treat the response as a valid discovery set because the payload still reports normal-looking results and brokerMetadata.nextStepHint suggests continuing the workflow.

Why this matters

The quick-discovery path is the default entry po

Read the thread · 2026-03-17 · closed · external user · 1 comment

[agentic-test] Known-item exact title lookup misses 'Attention Is All You Need' and dead-ends the documented smoke path

Summary

search_papers_match(query="Attention Is All You Need") returned a structured no-match payload during the smoke run, even though get_paper_details(paper_id="arXiv:1706.03762") still resolves the paper correctly. This breaks the repository's documented known-item golden path and creates an avoidable dead end for agents.

Why this matters

The public docs and the checked-in verifier both present search_papers_match as the first-step tool for messy or partial known-item lookup. A

Read the thread · 2026-03-17 · closed · external user · 0 comments

The remaining reports are on the project's issue tracker.