Hypotree MCP Server

MCP server: persistent, self-revising hypothesis-DAG for agentic R&D

Local serverstdioPython

What is the Hypotree MCP server?

Connect Hypotree to Claude, Cursor or any other MCP client and it stops being a tab you switch to. MCP server: persistent, self-revising hypothesis-DAG for agentic R&D. The hypotree mcp server is what makes that connection.

What the server does

A persistent, self-revising hypothesis DAG for agentic R&D — exposed as an MCP server.

  • Write-back belief revision — — an ATMS-style engine (de Kleer, 1986) that propagates evidence failures upstream through the dependency graph
  • Cascading prune — — invalidating a parent hypothesis instantly transitions its entire subtree to PRUNED. No tokens spent on dead branches
  • Exclusion-group inference — — confirming one member of a mutually exclusive group retires the rest as EXHAUSTED without probing them
  • Deduction by elimination — — last-man-standing: when all but one alternative in an exclusion group are refuted, the survivor is VERIFIED without a probe
  • Thompson Sampling navigation — — Beta-distribution sampling over the open frontier, giving bounded worst-case regret (no catastrophic lock-in)
  • Conflict resolution via differential ablation — — when an integration test fails but every component passes alone, the engine rebuilds the failing combination one swap at a time to pinpoint the culprit

Available tools

The toolset is worth reading before you wire it up, because it tells you what the integration is really for:

  • create_hypotheses — Create one or many nodes with parent_ids, exclusion_group, is_goal
  • get_next_targets — Thompson Sampling — returns the next hypothesis to test, under a lease
  • record_evidence — Record experiment results; triggers write-back propagation
  • generate_learning_path — What we learned, in order, and what it cost — separates conclusions an experiment paid for from ones the engine inferred free
  • get_workspace_info — Which belief state you are connected to and which layer chose it — start here when the graph is unexpectedly empty
  • update_status — Manually set node status (rarely needed — the engine does it)
  • get_dag_context — Get a subgraph view for the agent's context window
  • render_dag_map — Mermaid.js diagram of the current belief state
  • get_goal_status — Check whether the goal node is met
  • get_conflicts — List unresolved conflicts (integration failures)
  • suggest_discriminating_experiment — For a conflict, suggest the swap that separates the culprits
  • list_nodes — List/filter nodes by status, depth, or exclusion group

Installation

Installation goes through your MCP client rather than a global install: point it at hypotree on PyPI and it is fetched when the client starts. The copy-paste blocks for Claude Desktop, Claude Code and Cursor are further down this page.

Credentials and setup notes

Configuration is passed through the environment: HYPOTREE_WORKSPACE_ID. Treat anything key-shaped as a real credential — scope it to the minimum the server needs, and rotate it if it ever lands in a shared config.

Where it fits

This sits in the database access group, where several servers overlap in what they claim to do but differ sharply once you actually set them up. Hypotree's toolset — create_hypotheses, get_next_targets, record_evidence and 11 more — is a fair guide to whether it matches your workflow.

This entry was verified against Hypotree's own documentation before publication; SyncDev keeps the directory reviewed rather than auto-generated.

Worth knowing first

  • It runs with your machine's permissions. That is convenient and also the reason to think about what you point it at before you approve a tool call.
  • With 14 tools registered it takes up a noticeable share of the context window; turn it off in projects that never touch Hypotree.
  • Keep per-call confirmation enabled while you learn its behaviour; it is the cheapest safeguard you have.

Available tools

ToolWhat it does
create_hypothesesCreate one or many nodes with parent_ids, exclusion_group, is_goal
get_next_targetsThompson Sampling — returns the next hypothesis to test, under a lease
record_evidenceRecord experiment results; triggers write-back propagation
generate_learning_pathWhat we learned, in order, and what it cost — separates conclusions an experiment paid for from ones the engine inferred free
get_workspace_infoWhich belief state you are connected to and which layer chose it — start here when the graph is unexpectedly empty
update_statusManually set node status (rarely needed — the engine does it)
get_dag_contextGet a subgraph view for the agent's context window
render_dag_mapMermaid.js diagram of the current belief state
get_goal_statusCheck whether the goal node is met
get_conflictsList unresolved conflicts (integration failures)
suggest_discriminating_experimentFor a conflict, suggest the swap that separates the culprits
list_nodesList/filter nodes by status, depth, or exclusion group
get_evidence_historyFull evidence trail for a node
get_active_claimsList nodes with active leases

How to install the Hypotree MCP server

{
  "mcpServers": {
    "hypotree": {
      "command": "uvx",
      "args": ["hypotree"],
      "env": {
        "HYPOTREE_WORKSPACE_ID": "your-value"
      }
    }
  }
}

Add to claude_desktop_config.json, then restart Claude Desktop.

Configuration

VariableDescriptionRequired
HYPOTREE_WORKSPACE_IDConfiguration value read at startup.Optional

Example prompts to try

  • Use Hypotree to create hypotheses.
  • Use Hypotree to get next targets.
  • Use Hypotree to record evidence.

Frequently asked questions

It connects Hypotree to MCP-compatible AI assistants such as Claude and Cursor, exposing 14 tools (create_hypotheses, get_next_targets, record_evidence, and more) that the assistant can call on your behalf. Instead of copying data back and forth by hand, the assistant works with Hypotree directly.