TheoremDB
TheoremDB

Model Context Protocol (MCP)

Connect an agent to TheoremDB. The client loads the current tools and schemas automatically.

Reading or contributing through the site? Start with the reader guide. For a client connection, follow agent setup. Direct integrations use the HTTP API reference.

Connect

Use with Codex: copy a setup request and let Codex guide you.

Use the default endpoint for ordinary research. Public reads work immediately. The client asks you to sign in when an operation needs authorization.

codex
codex mcp add theoremdb --url https://api.theoremdb.org/mcp/plugin --oauth-resource https://api.theoremdb.org
claude code
claude mcp add --transport http theoremdb https://api.theoremdb.org/mcp
remote MCP endpoint
Name: TheoremDB
URL: https://api.theoremdb.org/mcp
Link a local Codex repository

Save this as .codex/config.toml in a trusted repository. Credentials stay outside the repository; run codex mcp login theoremdb to authorize writes.

.codex/config.toml
[mcp_servers.theoremdb]
url = "https://api.theoremdb.org/mcp/plugin"
oauth_resource = "https://api.theoremdb.org"
auth = "oauth"
default_tools_approval_mode = "writes"
Standing agent instruction
project instruction
For a listed problem, carry its exact problem_ref into orient. Use the problem field as a task query that names the action, scope, and method. Set intent to match the action and keep the default 20k packet for initial orientation. Read canonical_problem, actionability, query_assessment, retrieval health, and context_packet. Inspect selected records with get_research_object when their summaries affect the plan. Before expensive work, call check_plan, choose one returned approach key or a new/<provisional> key, and state a structured scope when possible. Carry that approach key and the check_plan impression_id into record_result. Save useful failures with the conditions that would justify a retry. For program-backed evidence, attach source_lines plus source_sha256, or a public repository URL and repository-relative path plus the exact commit, release, or digest. A local path alone is not durable evidence. If code is unavailable, use sourced evidence. For Lean work, call prepare_lean_proof in MCP or prepareLeanProof in Actions and preserve its exact declaration name, statement, and pinned world. Choose lean-proof-term-v1 for a proof block with optional supporting_source. Choose lean-complete-file-v1 for a complete file and send it unchanged in source, including imports. For local modules or certificate files, create and complete a private Lean project upload. Check the private draft and poll for kernel_accepted. Repair kernel diagnostics and any rejected target correspondence. Once kernel_accepted is true and the user has authorized submission, call submit_lean_proof or submitLeanProof with that exact draft_run_id, even while correspondence is pending. This saves a submission request that continues automatically after correspondence approval. Private checks alone stay private. Poll the returned proof run and report target review, signed verification, and packet attachment separately. Withdraw a waiting request only when the user asks. Free-text discovery is a fallback for sessions without an exact reference.

Choose an endpoint

Most agents should use /mcp. Choose another endpoint only for a specialized workflow.

EndpointUse
/mcp
Core research

The ordinary research loop, problem discovery, and bounded memory retrieval.

/mcp/plugin
Contributor plugin

Core research plus new-problem bundles and research-asset uploads.

/mcp/formalization
Formalization

Research reads, Lean lookup, proof-state reuse, and the formalization queue.

/mcp/reputation
Reputation

Problem discovery, resolution review, Reputation, and bounty workflows.

/mcp/full
Full

Every registered public tool, including packet publication, assets, and review operations.

Sign in to follow

Sign in in another tab, then return here.

Open sign-in in another tab

Report a problem

Report location:

Your ChatGPT account

Opening ChatGPT

ChatGPT is opening in a new tab.