Hexum
Server Details
Hosted MCP that shrinks coding-agent context before the model call; architecture checks without an LLM. Zero data retention.
- Status
- Healthy
- Uptime
- 100.0% over the last 24 hours
- Last Tested
- Transport
- Streamable HTTP · MCP 2024-11-05
- URL
- ChatGPT
- –
- Claude
- –
TDQS
Scored across 12 tools
Most tools have clearly distinct roles: loop lifecycle tools (start/status/tick/abort) are separate from code transformation tools (shrink/expand/splice). The only mild ambiguity is among hexum_check, hexum_review, and hexum_control_status, which all report drift/violations but differ by input scope (workspace, diff, session).
All names use snake_case and mostly follow an action-led pattern like check, expand, review, shrink, splice. The pattern is slightly inconsistent because control_status and loop_status are noun-led while the rest are verb-led, and the control_ prefix feels less uniform than the loop_ prefix.
Twelve tools is a well-scoped size for a coding-loop and architecture-enforcement server. Each tool maps to a meaningful operation: checking, reviewing, editing, shrinking, expanding, loop control, and skill management.
The tool surface covers the full workflow: check architecture drift, review diffs, transform code via shrink/expand/splice, run and control autonomous loops, synthesize architectural fixes, and manage the reply skill. No critical dead-end is apparent; loop tools provide start/status/tick/abort, and synthesize returns patches the agent can apply.
Available Tools
12 toolshexum_checkAInspect
Check the workspace against architecture.hexum and report layer/import drift. Deterministic; no LLM. Report may include skill (lite|full|ultra). If skill is on and SKILL.md is missing, call hexum_skill.
Inputs: files, paths, workspace.
hexum_control_statusAInspect
Drift report for this MCP session: score 0-100, health healthy|drifting|stuck_in_loop, file churn, recent violations. Call when the agent is repeating edits or failing the same check.
Inputs: no named parameters.
hexum_control_synthesizeAInspect
From this session's violations, synthesize architecture.hexum forbid lines and SKILL.md directives the agent should write. Does not write the files; returns patches.
Inputs: current_config.
hexum_expandAInspect
Expand one skeletonized function, class, or method: return its full source.
Inputs: file_path (required), symbol_name (required).
hexum_loop_abortAInspect
Stop a running loop. Returns the transcript summary. In-memory only; nothing is stored.
Inputs: loop_id (required), reason.
hexum_loop_startBInspect
Start an autonomous build loop for a goal. Returns loop_id and the next instruction (shrink, edit, check, test, tick). The model still writes code; Hexum gates each turn.
Inputs: goal (required), files, max_turns, test_command.
hexum_loop_statusAInspect
Read loop state: turns, gates, tokens, health. Does not advance the loop.
Inputs: loop_id (required).
hexum_loop_tickAInspect
Advance a Hexum loop one turn. Pass compile_output and test_output (and updated files). Runs hexum_check, condenses logs, scores drift, and returns the next instruction or a terminal status (passed|aborted).
Inputs: loop_id (required), files, notes, test_output, +2 more.
hexum_reviewAInspect
Review a git unified diff against architecture.hexum. Deterministic doctor plus token delta. Returns approved|changes_requested, violations, and a GitHub-ready markdown review. Prefer this over asking the model to review architecture.
Inputs: diff, files, config, workspace.
hexum_shrinkAInspect
Shrink workspace files into AST-valid skeletons (hot / call-chain / peripheral / manifest). Returns compressed sources plus token stats. Prefer workspace= when the server can see the disk; otherwise pass files. Default budget 8000. Pass since_hash from the prior snap to skip unchanged files. Use this before reading a repo. If architecture.hexum has a skill line and SKILL.md is not installed, call hexum_skill first.
Inputs: files, paths, stderr, workspace, +4 more.
hexum_skillAInspect
Install or switch the Hexum reply skill. Reads skill lite|full|ultra|off from architecture.hexum, or set intensity in this call. Returns SKILL.md body plus write path. If the file is missing, write body to path then follow it. Call this before other Hexum tools when the skill is not installed.
Inputs: agent, files, intensity, workspace.
hexum_spliceCInspect
Small AST-valid write. Pass symbol_name + new_implementation to replace a symbol, or line + new_source to replace the node at a line.
Inputs: file_path (required), line, column, node_kind, +3 more.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Related MCP Servers
- AlicenseNot gradedqualityCmaintenanceRemote MCP server that lets AI assistants read tracked competitive intelligence — competitor news, reviews, website changes, social posts and ads — and act on it by managing projects, competitors, battle cards, reports and canvases. Every answer carries a source and a date, and write actions only publish when explicitly shared.MIT
- AlicenseCqualityCmaintenanceCompetitor Monitor AI - MCP server providing AI-powered tools and automation by MEOK AI Labs117 npm46 PyPIMIT
- AlicenseAqualityDmaintenanceEnables brand visibility monitoring across major AI platforms like ChatGPT, Claude, Gemini, and Perplexity. It allows users to track visibility scores, analyze competitor data, and receive actionable insights to improve AI-generated brand recommendations.169 npm1MIT

industrylens-mcpofficial
AlicenseNot gradedqualityCmaintenanceBrowse IndustryLens's published competitive-intelligence reports and head-to-head competitor comparisons from any AI agent — real, source-backed data.MIT
Glama MCP Gateway
Add one secure layer between your agents and this server.