# Reported issues for DevGlobe Pod holds 16 of 20 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 [DevGlobe](/mcp/devglobe). ## Most discussed ### Agent analytics: measure privacy-safe MCP adoption and outcomes ### Problem We lack telemetry on how agents interact with the DevGlobe MCP server, preventing us from optimization and tracking growth safely. ### Acceptance Criteria - Track aggregate metrics: initialize, ools-list, ool-call, errors, latency, and returned result counts. - Track introduction conversion outcomes. - Ensure no raw private content is captured (privacy-safe). - Document data retention policy. - Provide dashboard queries and documentation. ### Dependencies - None. [Read the thread](https://github.com/sajeetharan/devglobe/issues/201) · 2026-08-18 · open · 4 comments ### Epic: Make DevGlobe discoverable and reusable by AI agents ### Problem DevGlobe needs to become the premier discoverable and reusable platform for AI agents. We must implement standardized MCP capabilities, authentication, onboarding portals, analytics, and workflow recipes, then publish them to registries. ### Implementation Status Child Issues: - [ ] #196 Agent discovery: publish DevGlobe to MCP registries and client directories - https://github.com/sajeetharan/devglobe/issues/196 - [ ] #197 Agent onboarding: build an Add to Agent setup hub - https: [Read the thread](https://github.com/sajeetharan/devglobe/issues/203) · 2026-08-18 · open · 2 comments ### MCP contract: add tool annotations and standardized errors ### Problem The MCP server doesn't provide stable, annotated hints for tool discovery or standardized error guidelines, making automated retries and agent-side error handling difficult. ### Acceptance Criteria - Provide read-only annotations for discovery tools. - Provide state-changing annotations for introduction tools. - Implement stable error codes and structured categories. - Document and implement retry/rate-limit guidance for client agents. - Tests and developer documentation. ### Depen [Read the thread](https://github.com/sajeetharan/devglobe/issues/199) · 2026-08-18 · closed · 2 comments ### MCP contract: add structured outputs and explainable discovery results ### Problem AI agents require clear, structured outputs and explainable matches when performing developer discovery searches on DevGlobe to build trust in recommendations. ### Acceptance Criteria - Add outputSchema and structuredContent to all public discovery tools. - Search result items must include: profileUrl, whyMatched, public evidence, freshness, agent availability, and methodology disclaimer. - Maintain backward-compatible text content outputs. - Comprehensive test coverage. ### Depend [Read the thread](https://github.com/sajeetharan/devglobe/issues/198) · 2026-08-18 · closed · 2 comments ### Agent onboarding: build an Add to Agent setup hub ### Problem Developers and agents have no dedicated, user-friendly portal to learn how to integrate and connect DevGlobe's MCP tools to their environments. ### Acceptance Criteria - Build a dedicated, accessible, and responsive /agents page. - Visual configurations and setup copy controls for VS Code, Claude Desktop, Cursor, and generic Streamable HTTP. - Include clear links to the endpoint, server card, Agent Skill, and auth docs. ### Dependencies - None. [Read the thread](https://github.com/sajeetharan/devglobe/issues/197) · 2026-08-18 · closed · 2 comments ### Agent discovery: publish DevGlobe to MCP registries and client directories ### Problem DevGlobe's hosted MCP service needs to be discoverable by AI agents and developers. We lack official presence in major MCP registries and client directories, which limits our adoption and reach. ### Acceptance Criteria - Official MCP Registry submission metadata is documented and ready. - Relevant client directories (e.g., Claude Desktop, Cursor, VS Code) are evaluated and submitted to. - Use canonical endpoint and server card. - Submission status is recorded and tracked. ### Depen [Read the thread](https://github.com/sajeetharan/devglobe/issues/196) · 2026-08-18 · open · 1 comment ### feat: expose DevGlobe as a hosted MCP service for AI agents ## Problem DevGlobe already supports consent-aware developer discovery and introductions through a local stdio MCP server, but agent owners must clone the repository and run Node locally. This prevents agents from using DevGlobe as an internet-native service. ## Proposal Expose a hosted, stateless MCP Streamable HTTP endpoint at `/mcp` so MCP-compatible agents can discover developers and request consent-gated introductions directly through `https://www.devglobe.dev`. ## Agent tools - `search_d [Read the thread](https://github.com/sajeetharan/devglobe/issues/139) · 2026-08-13 · open · 1 comment ### Public MCP service status and capability page ## Problem There is no public, simple way to understand the status, version, and capabilities of the public MCP service, which makes adoption and troubleshooting harder for clients. ## Proposed solution Publish a public MCP service status and capability page that includes uptime, version, tools, prompts, resources, client setup guidance, and aggregate reliability information without exposing identity or user-specific data. ## Acceptance criteria - Public status and capability page. - Uptime, v [Read the thread](https://github.com/sajeetharan/devglobe/issues/332) · 2026-08-30 · open · 0 comments ## Most recent ### Saved-search MCP tools ## Problem MCP clients need a safe way to save and run searches, but anonymous persistence and unbounded schema can create privacy and complexity problems. ## Proposed solution Add authenticated saved-search MCP tools that support list, create, delete, and run actions using the #166 model, with no anonymous persistence and bounded schema validation in the tool contract. ## Acceptance criteria - Authenticated list/create/delete/run behavior. - No anonymous persistence. - Bounded schema and vali [Read the thread](https://github.com/sajeetharan/devglobe/issues/330) · 2026-08-30 · open · 0 comments ### Stable cursor pagination for developer search ## Problem Developer search lacks a stable and deterministic pagination mechanism. Clients need a predictable cursor-based API and MCP contract that can survive reordering and invalid or expired continuations without exposing unstable internal indexes. ## Proposed solution Implement stable cursor pagination for developer search which uses an opaque continuation token, deterministic ordering, and bounded page sizes. The contract should define max 20 results, invalid cursor handling, and expirati [Read the thread](https://github.com/sajeetharan/devglobe/issues/329) · 2026-08-30 · open · 0 comments ### Project-first developer discovery ## Problem Users begin from a repository, not a person. The product currently has no first-class path to understand whether a public codebase matches a developer profile or to explain why a person is likely relevant to a repository. The system must accept a public GitHub owner/repo, resolve repository language, topics, and description, and then compare it against indexed public developers with transparent evidence instead of opaque ranking. ## Proposed solution Add a repository-first developer [Read the thread](https://github.com/sajeetharan/devglobe/issues/325) · 2026-08-30 · closed · 0 comments ### Feature: Live Agent Scout discovery workspace ## Problem search is a one-shot result list and the MCP agent experience is detached from the web experience. ## Proposed Solution a visible, user-controlled scout workspace where a natural-language mission is converted into search criteria and progressively builds a shortlist with evidence-backed inclusion/exclusion explanations. ## MVP Scope mission composer with structured interpretation preview; user must confirm criteria before execution; progress states; candidate stream; inspect evidenc [Read the thread](https://github.com/sajeetharan/devglobe/issues/212) · 2026-08-18 · open · 0 comments ### Agent authentication: design OAuth 2.1 onboarding for introduction tools ### Problem To allow agents to request/perform state-changing introduction tools securely on behalf of users, we need a robust OAuth 2.1 protocol design instead of static pre-issued keys. Public searches must remain anonymous. ### Acceptance Criteria - Complete threat model and security design document. - Detail protected resource metadata and authorization server decision. - Define delegated scopes for introduction actions. - Implement token rotation and revocation. - Define transition/migrati [Read the thread](https://github.com/sajeetharan/devglobe/issues/200) · 2026-08-18 · open · 0 comments ### Build comprehensive feature documentation and deploy it to GitHub Pages ## Summary Create a dedicated, public documentation site for DevGlobe that explains the product's human-facing features and its agent-ready surfaces. Deploy the static documentation to GitHub Pages while keeping the dynamic Next.js application on Vercel. ## Goals - Give developers, users, and AI-agent builders one navigable source of truth. - Document current behavior from the implementation, not aspirational features. - Make MCP, Agent Skills, OpenAPI, WebMCP, machine-readable discovery, authe [Read the thread](https://github.com/sajeetharan/devglobe/issues/188) · 2026-08-17 · closed · 0 comments ### Visualize the consent-based Agent Network and connection lifecycle ## Problem DevGlobe now supports consent-gated MCP introductions, but the interaction is mostly invisible outside the private request inbox. This makes an important differentiator difficult for developers and visitors to understand, trust, and share. ## Goal Visualize the developer-agent connection lifecycle in a privacy-safe way that helps developers understand the value, opt in, and share their agent-ready profile. ## Proposed MVP ### Agent Network view Add an **Agent Network** view alongsi [Read the thread](https://github.com/sajeetharan/devglobe/issues/138) · 2026-08-13 · open · 0 comments ### Add consent-based AI tools and agent collaboration profiles ## Problem DevGlobe describes an agent-ready developer discovery layer, but developer profiles cannot currently declare which AI coding agents they use or whether they accept agent-initiated collaboration requests. GitHub data cannot reliably establish private tool usage, so this must be consent-based and self-declared. ## Proposed MVP - Allow a signed-in developer with a claimed profile to manage an AI collaboration profile. - Capture selected AI tools, usage level, collaboration availability, [Read the thread](https://github.com/sajeetharan/devglobe/issues/135) · 2026-08-13 · closed · 0 comments The remaining reports are on [the project's issue tracker](https://github.com/sajeetharan/devglobe/issues).