Universal Research MCP

Install from PyPI · Read the documentation · Report an issue

Universal Research MCP is a provenance-first, append-only research operations framework and governed MCP. It records plans, approvals, observations, claims, failures, amendments, and contributions with traceable sources. Canonical JSONL is authoritative; SQLite search indexes are verified, replaceable derived views.

python -m pip install --upgrade universal-research-mcp

After installation, use either the full universal-research command or the short urmcp alias. The PyPI distribution name remains universal-research-mcp, so users can find the authoritative package and its release history in one place.

Supported integration (0.5.3): Codex only. Codex owns model selection, agent sessions, tool execution, approvals, and GUI presentation. Ollama, OpenAI API, Anthropic API, Moonshot/Kimi, Claude Code, OpenCode, and OpenClaw are not supported or invoked by this release.

The repository can contain provider or runtime prototypes for future research. They are not exposed through the default CLI, Codex plugin, MCP tools, or package entry points, and are not a compatibility promise.

Quick start

Secure Codex/Docker harness and governed-result promotion

The secure harness keeps Codex on the host and runs only sealed test, build, and experiment recipes in offline, resource-bounded Docker workers. The original project is never mounted into a worker, visualization is disabled, and edits stay quarantined until their exact diff hash is approved. See the secure harness guide. A canonical result declared as a benchmark or final_review must carry an exact secure-harness attestation; ordinary Codex work cannot be promoted under those labels without one. This preview does not claim to be a custom Codex runtime or a complete multi-provider agent scheduler.

Python 3.11 or newer is required.

python -m pip install universal-research-mcp
universal-research init ./my-research
universal-research serve --root ./my-research

When run directly in a terminal, serve shows its startup phases on the terminal's status stream—for example index check, staged rebuild, verification, and 100% ready for MCP requests. This never writes to the MCP protocol stream. It is enabled by default for an interactive terminal and disabled by default for a non-interactive MCP host. Override it with --startup-progress or --no-startup-progress.

init creates an independent empty canonical source registry and a verified FTS5 lexical index. It reports the semantic index as missing; it does not silently create an empty semantic database, download a model, or contact an embedding API. A lexical refresh that does not alter canonical JSONL is staged, validated, and atomically replaced. Select the research root with --root or UNIVERSAL_RESEARCH_ROOT.

Optional semantic and hybrid candidate retrieval

After canonical records exist, semantic retrieval is an explicit project-level choice. The offline demo backend creates no download, GPU, credential, or network request, and is useful for checking the full semantic/hybrid workflow:

universal-research semantic configure --backend demo --root ./my-research
universal-research semantic build --root ./my-research
universal-research semantic status --root ./my-research

The demo identifies itself as backend_class=deterministic_demo and trained_embedding_model=false. It is a reproducible retrieval fixture, not a learned embedding model or a quality claim. For a pinned model already present on disk, configure --backend local --model-path /absolute/path/to/model; the command never downloads a snapshot. A configured semantic view is rebuilt only when --auto-refresh was explicitly selected; otherwise canonical writes leave it stale and print a recovery command. Remote embedding remains unsupported by the public MCP and is never invoked by serve.

Local SentenceTransformer setup

pip install universal-research-mcp never creates a virtual environment or downloads a model. To install the optional library dependencies into your current Python environment, use pip install "universal-research-mcp[semantic]". To keep a research project self-contained instead, use the built-in guided setup. It offers a reviewed catalogue of ten local SentenceTransformer models; the Korean/multilingual default is intfloat/multilingual-e5-base (768 dimensions). The catalogue includes compact, balanced, and large English and multilingual alternatives.

universal-research semantic models
universal-research semantic setup --root ./my-research \
  --model intfloat/multilingual-e5-base --device cuda

The first command only prints a plan and a plan_sha256. With the default environment manager, setup uses Conda when the conda executable is available and otherwise selects Python venv. It does not create either one yet. Review the planned environment path, package version, model revision, device, and network use; then repeat the exact hash to authorize the mutating operation:

universal-research semantic setup --root ./my-research \
  --model intfloat/multilingual-e5-base --device cuda --execute \
  --confirm-plan-sha256 <plan_sha256>

The confirmed step creates or explicitly reuses ./my-research/.universal-research/semantic-env, installs the exact package version with its semantic extra, downloads only the selected reviewed model to ./my-research/.universal-research/models/, and writes the local semantic configuration. It does not build an index or start a model. Its result gives the environment-specific semantic build and serve commands. Point Codex's MCP server command to that returned environment-specific universal-research executable only after reviewing the host configuration change.

Use --revision <immutable-model-commit> instead of the default main when reproducibility matters. Setup will never replace an existing environment or model directory unless you regenerate the plan with --reuse-existing and confirm its new hash. The system uses trust_local_model_code=false; arbitrary repository IDs, raw model code, external provider APIs, and automatic model downloads are not accepted through this flow.

See the full canonical input tutorial for a source → human approval → guarded record append → lexical search workflow.

Declarative research profiles

For a repeatable project policy, use a JSON profile. It can declare whether candidate retrieval is lexical, semantic, or hybrid; select the already-present offline demo or local-GPU backend; bound source discovery to documentation, source code, build definitions, and configuration; and record approved future provider routes. It is not an agent runner or credential store.

The optional candidate_backend policy controls lexical candidate ordering and hybrid fusion without changing the canonical ledger. "universal" (the default) searches exact source passages before event summaries and uses the current 0.45/0.55 hybrid weights. "event_first" preserves the predecessor Research Memory policy: event-summary FTS first and equal-weight RRF. It is implemented inside this package against Universal's own current derived views; it never opens a predecessor database, launches its watcher, or executes an external project script. The small matched evaluation that motivated this option is not evidence that either policy is universally better.

universal-research profile template > research-profile.json
universal-research profile validate research-profile.json --root ./my-research
# Review the returned profile_sha256, then repeat it exactly:
universal-research profile apply research-profile.json --root ./my-research \
  --confirm-profile-sha256 <profile_sha256>
universal-research profile status --root ./my-research

The template is lexical-only, network-disabled, and keeps the documented source categories available for a later explicit registration step. To enable the offline reproducible semantic fixture, change the profile's retrieval section to:

{
  "mode": "adaptive",
  "candidate_backend": "event_first",
  "semantic_backend": {
    "kind": "demo",
    "dimensions": 256,
    "auto_refresh": false
  }
}

Then run universal-research semantic build --root ./my-research. adaptive uses lexical retrieval for explicit file paths, command flags, and code identifiers; it uses semantic retrieval for ordinary research questions when the configured semantic index is current. If that semantic view is unavailable or returns no candidates, it reports a lexical fallback in the response rather than implying that semantic retrieval ran. It does not use an unvalidated score threshold or claim a universal accuracy improvement. For a local GPU model already on disk, use kind: "local", a model_path, and device: "cuda"; applying the profile never downloads or loads that model. The public MCP exposes the profile's status but cannot call a declared OpenAI, Anthropic, Ollama, or other provider route. A remote route requires an explicit network_enabled: true declaration and an env:NAME credential reference; the profile never contains a raw API key and the public MCP still does not read it or make the call.

Profiles can select only the package's registered Codex Skills (research-governance, research-workflow). You can author and install another Codex Skill, but a profile cannot create or activate it by itself. Add a new Skill through the reviewed plugin/release path, then add its fixed ID to the runtime registry; this prevents a retrieved document or an agent from creating its own authority and executing it.

First searchable input

The management CLI remains available for explicit administration. In 0.5.0, the same universal_research MCP can also perform a bounded two-step ingest:

universal-research init ./my-research
# Create ./my-research/docs/note.md yourself, then register that immutable path.
universal-research source register docs/note.md --root ./my-research \
  --source-id src_note_v1 --source-type markdown
universal-research record template > protocol.json
universal-research record validate protocol.json --root ./my-research

For a governed record, first append a human record_kind=approval record with record approve --confirm <record-id>. Then have Codex call research_prepare_ingest with the record, that approval ID, and any new project-contained source registrations. It stores an immutable non-canonical pending draft and returns its ID and SHA-256. After reviewing the intended append, issue a one-time local receipt outside the MCP process:

universal-research ingest approve --root ./my-research \
  --draft-id ingest_... --draft-sha256 <draft-sha256> \
  --confirm-draft-sha256 <draft-sha256> \
  --expires-at 2026-08-15T00:00:00+00:00

Then allow the host-visible mutating research_commit_ingest call with only that exact draft ID, hash, and returned receipt ID.

Commit does not accept a replacement record body or approved=true flag. It requires the existing human scope approval plus a signed, one-time receipt stored outside the project; it checks the receipt, draft hash, canonical head, and source files, consumes the receipt and draft once, then appends the record and refreshes lexical plus an explicitly configured auto-refresh semantic index. A derived-index failure never hides a successful canonical append; it is reported as stale/partial in the commit audit result. Registered paths are immutable revisions: editing or re-registering one is rejected; register a new project-contained path instead.

The default universal_research MCP provides:

• candidate search and event/hash-bound source re-verification;
• deterministic claim eligibility receipts for material results, comparisons, causal statements, release decisions, and explicitly material facts;
• an eleven-role governance contract and Codex dispatch-manifest preparation;
• scope/cost preflight, deterministic operation-gate, and failure-tombstone preparation; and
• lexical, semantic, and RRF hybrid candidate retrieval plus derived-index status.

It never creates human approval records, treats an agent assertion as approval, silently updates canonical history, invokes a remote API, downloads a model, or starts a benchmark or daemon. The MCP protocol does not carry a portable, signed proof of a specific Codex approval click. The receipt authority therefore uses a separate user-state signing key and explicit CLI confirmation; the server also enforces exact draft/state and human-scope checks. Query-time semantic search uses only the explicitly configured offline backend; an unconfigured or stale semantic view fails closed. Provider fallback is future work.

Evidence flow

canonical JSONL → staged/verified SQLite candidate → memory_fetch_evidence
with event_id + expected_sha256 → current-hash check → memory_gate_claim →
eligible or blocked material claim

memory_search_candidates defaults to mode="configured", which resolves the explicit project profile (or preserves lexical behavior when no profile exists). It also resolves candidate_backend="configured"; callers may explicitly use "universal" or "event_first" for a bounded comparison. Before candidates are returned, a deterministic identity gate requires every event/path/hash/line locator to exist in the current canonical projection. A backend mismatch fails closed rather than becoming a search result. It also accepts explicit "lexical", "semantic", "hybrid", and "adaptive" modes. Adaptive mode uses a deterministic syntax rule for the lexical fast path and otherwise tries the configured semantic view; every response reports the selected mode and routing reason. Hybrid search independently ranks lexical and semantic candidates and fuses ranks with reciprocal-rank fusion (0.45 lexical / 0.55 semantic, k=60). Every result remains candidate_only=true; semantic similarity cannot establish a fact, cause, comparison, or release claim. A semantic passage may choose the returned path and exact line range, but it still must go through memory_fetch_evidence and memory_gate_claim.

Search results are candidates, not evidence. For a consequential conclusion, retain the evidence fetch's event ID, path, line range, expected and current hashes, and integrity_status. Files not registered in the index cannot be fetched. Search, latest, and evidence fetch fail closed while the lexical index is stale and require universal-research index ensure --kind lexical. When a file's current hash differs, content is withheld by default; only the explicit diagnostic allow_mismatched_content=true response includes current content.

memory_gate_claim is a deterministic, fail-closed receipt for material claims. It re-fetches each supplied exact evidence reference and accepts only a current registered revision. Release, comparison, and causal claims require two distinct verified records; a material factual or result claim requires one. Routine lookups are not gated. The gate checks evidence eligibility, not the scientific truth of prose, and it does not invoke a model or write the ledger.

Governed Codex agents

scope_and_cost_governor runs before plan approval in every mode. It assesses necessity, a bounded time estimate, work units, difficulty, compute/network cost, and evidence; it does not approve execution or kill a process. The common operation gate performs a declarative preflight bound to an approved scope_hash and always returns execution_authorized=false. At the real tool boundary, the Codex host must compare the gate hash with a closed, action-specific argument envelope before deciding whether to execute.

The plugin prepares a role-specific task packet, hash-bound scope receipt, and Codex dispatch manifest over one evidence snapshot. A manifest does not start an agent. Codex may create native subagents and decide their sessions, model, parallelism, and GUI surface under host permissions and the user's entitlement. The plugin does not bypass a Codex subscription by routing work to a paid API.

The default failure policy is blocking_only + ask + redacted. A minimum failure tombstone is always retained, while detail retention is configurable as full | metadata_only | ask and full | redacted | hashes_only. There is no unrecorded off mode.

Host visualization is off by default. It requires both explicit user opt-in and a separately approved capability scope/plan reference. Permission to generate a normal data plot does not imply permission to invoke a host visualization skill.

Token accounting uses only exact counts supplied by a provider or host. If the host does not expose an exact count for commands, code generation, Skills, or visualization, that category is unavailable, not zero or an estimate.

Benchmark status

The repository provides paired A/B protocols, fixture contracts, and scoring contracts. No confirmatory live A/B result is available.

The graphic reports two public, exploratory live diagnostics using synthetic sources, gpt-5.6-terra, low reasoning effort, and one run per condition. On four ordinary single-source tasks, both conditions achieved 4/4 factual answers with evidence-line citations; the MCP condition used 2.26× input tokens and 2.02× wall latency. In the separate six-trial post-index source-mutation diagnostic, direct filesystem retrieval accepted the changed source as verified evidence in 6/6 trials, while MCP-gated retrieval accepted it in 0/6 and abstained correctly in 6/6. This is bounded evidence for stale-source protection, not evidence of general hallucination reduction, research-quality improvement, or model superiority. See the machine-readable directional source diagnostic and the claim-safety pilot.

The narrower public development protocol integrity-claim-gate-v1 has now been executed once: 24 public synthetic tasks × 4 conditions = 96 participant runs, with a separate condition-blinded evaluator. The observed MCP + Claim Gate condition had 2/18 unsafe material assertions on fault tasks (11.1%), versus 4/18 (22.2%) for direct filesystem; clean supported-claim coverage was 100.0% versus 66.7%. It cost 1.55× mean execution tokens and 1.61× mean latency. The paired 95% interval for the unsafe-assertion difference includes zero, so this is not proof of an effect. It is a development-sample signal plus a measured burden and known semantic failure modes. See the complete development result.

Future runs must use the same model, prompt, tasks, permissions, and source snapshot within each pair; retain failures; use paired repetitions; and preserve raw host telemetry. The only reportable public metrics are citation correctness, unsupported-claim rate, evidence retrieval success, total tokens, tool calls, latency, cost, and paired differences. Missing telemetry is unavailable, never inferred. See the benchmark disclosure.

Boundaries and data authority

Reference projects are read-only design inputs. Universal Research does not create, modify, delete, move, or append their files, event logs, databases, results, or session records. Their embedding databases may be read only to understand a schema, metadata, or adapter design. The new project never copies or shares a reference runtime database.

Semantic/hybrid retrieval is a hardened successor to the public Evidence-First Research Memory predecessor prototype: it retains separate lexical/semantic ranking and source-range candidates, while adding registered-file SHA-256 re-verification, stale fail-closed behavior, and deterministic claim gating.

1. data/events/ JSONL is the canonical event ledger.
2. data/index/ SQLite and any embedding index are rebuildable derived views.
3. Original sources and artifacts verify candidate results.
4. Embedding similarity alone cannot establish a fact, cause, or performance claim.

Project-local working notes, when used, are not canonical records and are not distributed with the package.

Architecture

universal_research_mcp/
  plugin/          Codex plugin and Skills
  core/            canonical-input and record contracts
  governance/      fixed-role governance contracts and validators
  integrations/    host-specific dispatch adapters
  data/events/     independent canonical append-only events
  data/index/      independent derived lexical/dense indexes
  docs/            specifications and usage documentation

The core schema defines immutable records and typed relations. Study-type and domain packs may add restrictions but may not relax core policy. A project profile selects paths, adapters, and reference boundaries. The MCP normally returns candidate metadata and exact source evidence; its only canonical-write surface is the host-visible, hash-bound two-step ingest boundary.

Current support and next steps

This release supports the Codex plugin, local lexical lifecycle, source-grounded evidence fetch, guarded canonical ingestion, eleven-role governance, and non-executing Codex dispatch preparation. Installation, MCP startup, and CI do not invoke an API, local model, model download, benchmark, or background watcher.

The 0.4.0 namespace is intentionally breaking: public Python modules live under universal_research_mcp.*; legacy general top-level package imports are not provided as shims. Independently reviewed adapters for other hosts remain separate work.