Havoptic MCP¶

The tools an engineering team codes with now ship faster than anyone can read. The upgrade that breaks the build is usually documented, in a changelog nobody opened.

Server type: Remote (Streamable HTTP), keyless
Endpoint:    https://mcp.havoptic.com/mcp
Auth:        None, read-only
Server:      havoptic 2.1.0, protocol 2025-06-18
Tools:       15 at tools/list, 4 of them deprecated aliases
Tracking:    Claude Code, Codex CLI, Cursor, Gemini CLI, Kiro CLI, Copilot CLI, Devin Desktop, agent SDKs, inference runtimes
Pricing:     Free

Why This Matters for Operators¶

A team that standardises on Claude Code, Codex and Cursor is running three release trains it does not control. A flag gets renamed, a hook changes shape, a permission model tightens, and the first sign is a broken workflow on a Monday. The information exists, inside three changelogs, in prose, released weekly.

This server turns the changelogs into typed questions: what changed since this date, what breaks between these two versions, which tool shipped a capability first and how far behind the others are, and how fast each vendor ships at all. whats_changed_since accepts relative windows like 7d and 30d and groups the answer by tool with breaking changes first, which is the shape of a Monday morning question rather than the shape of a changelog.

It is adjacent to the catalogue's Versionly entry and distinct from it: Versionly watches the vendor changelogs a company's own code depends on, from a GitHub App against selected repos. Havoptic tracks the assistant and agent tooling itself, across vendors, with feature-adoption timing and an editorial comparison.

What It Returns, Verified¶

A bare initialize on https://mcp.havoptic.com/mcp returned HTTP 200 with no credential as havoptic 2.1.0 on protocol 2025-06-18. tools/list returned fifteen tools: eleven current and four kept as deprecated aliases that point at their successors, which is itself a useful signal about how the vendor handles its own surface.

The eleven current tools:

list_tools is the catalogue of tracked products: AI coding tools plus model and API platforms, agent SDKs and inference runtimes, with release counts.
get_latest_releases returns the newest releases across one or all tracked tools, with a since and until window.
whats_changed_since returns typed changes, breaking, security, features and fixes, grouped by tool with breaking changes first.
search_changes is full-text search across typed change items and legacy release notes, for a flag, an error string, a config key or a feature name.
get_release takes a release id such as claude-code@2.1.183 or a tool plus version, with a full-detail mode.
get_upgrade_notes returns everything between two versions of one tool, including an explicit coverage gap where legacy releases carry no typed changes.
get_breaking_changes lists breaking changes and deprecations across tools in a window, 30 days by default.
compare_tools compares two to four tools on release cadence, feature-matrix differences and recent changes.
feature_adoption answers who shipped a cross-tool feature first, such as MCP support, subagents, planning mode, hooks or checkpoints, and the lag in days for the rest.
cadence_stats gives release counts over 7, 30 and 90 days, average days between releases and the last release.

Limits Worth Recording¶

Read-only, so it informs decisions and changes nothing. The typed changes and the coverage gap matter more than the release count: the vendor states that legacy releases without typed changes are counted separately rather than silently treated as change-free, which is the honest shape and worth checking before anyone concludes a version was quiet. This tracks the assistant and coding-agent tooling layer, not the wider SDK and cloud changelog space, so Versionly remains the entry for a company's own dependency surface.

FAQ¶

Do I need an API key?¶

No. All fifteen tools answer without a credential, which is how this entry was verified.

Which tools are tracked?¶

AI coding tools including Claude Code, Codex CLI, Cursor, Gemini CLI, Kiro CLI, GitHub Copilot CLI and Devin Desktop, plus model and API platforms, agent SDKs and inference runtimes.

Can it tell me what breaks when I upgrade?¶

Yes. get_upgrade_notes returns everything between two versions of one tool with breaking changes, deprecations, security fixes, features and fixes, and get_breaking_changes does the same across tools in a window.

Does it tell me which tool shipped a feature first?¶

Yes. feature_adoption answers who shipped a cross-tool feature first and how many days behind the others were, and compare_tools compares two to four tools including cadence.

How is this different from Versionly in this catalogue?¶

Versionly watches the vendor changelogs your repositories depend on, through a GitHub App against selected repos. This tracks the coding assistants and agent tooling themselves across vendors, including cross-tool feature timing.

Integration with CorpusIQ¶

CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.

This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.

See what CorpusIQ connects