# codex-security

OpenAI's Codex Security CLI and TypeScript SDK for finding, validating, and fixing security vulnerabilities. npm: https://www.npmjs.com/package/@openai/codex-security

- **Kind:** MCP server
- **Owner:** openai
- **Source:** https://github.com/openai/codex-security
- **Runtime:** node
- **License:** Apache-2.0
- **Transports:** stdio
- **Pinned commit:** `64bb3607067f0530b06c3be4b687d4d0bfd8dc9a`
- **Page:** https://forefy.com/mcps/c7d55997-0eea-5491-a69b-938f7b4e9421
- **API (JSON):** https://forefy.com/api/asr/mcps/c7d55997-0eea-5491-a69b-938f7b4e9421

## Install

Ask your agent: `Install the codex-security MCP server securely from: https://forefy.com/mcps/c7d55997-0eea-5491-a69b-938f7b4e9421 and register it with my agent, prompting me for GH_TOKEN, TENABLE_ACCESS_KEY, TENABLE_SECRET_KEY`

## Tools (53)

- `get_codex_security_daybreak_access`: Check this ChatGPT account
- `start_codex_security_standard_scan`: Headless and CLI only. Start or rejoin a Standard security scan. Do not use for desktop scans, Review changes, Deep Scan, or an existing SDK-owned scan. Use th…
- `start_codex_security_prompt_only_scan`: Start or rejoin a Standard or diff Codex Security scan from its owning conversation. Use the returned authoritative scanId and scanDir. Standard and diff scans…
- `request_codex_security_user_input`: Fallback for interactive Codex Security workflows when the host-native request_user_input tool is unavailable. Presents one to three non-sensitive multiple-cho…
- `open_codex_security_workspace`: App-only. Create a native Codex Security workspace with the target and requested standard, diff, or deep mode, or reopen one owned by this thread by passing on…
- `inspect_codex_security_target`: App-only. Validate a local target directory and derive its display and Git metadata without saving setup.
- `inspect_codex_security_setup`: Resolve and validate the local target, scope, mode, and exact Git change set without saving setup. Deep Scan returns effective limits and their sources before…
- `submit_codex_security_setup`: App-only. Validate and save bounded target, scope, mode, and optional context selections.
- `start_codex_security_scan`: App-only. Create a scan record and its local artifact directory before Codex analysis begins.
- `start_codex_security_deep_scan`: Run or rejoin independent Standard security scans and return their deduplicated findings. Pass scanId and its handoffClaimToken to resume, or targetPath to sta…
- `cancel_codex_security_scan`: Stop a running scan from its owning Codex thread, prevent further progress or completion updates, and cancel any active deterministic Deep Scan SDK workers.
- `cancel_codex_security_scan_from_app`: App-only. Stop a running scan from the native Codex Security workbench, prevent further progress or completion updates, and cancel any active deterministic Dee…
- `recover_codex_security_scan_results`: App-only. Explicitly validate and republish retained checkpoints for one stopped scan, updating its artifacts, finding index, and counts.
- `get_codex_security_scan`: App-only. Read plugin-owned scan state for native Security monitoring without claiming a pending Codex handoff.
- `list_codex_security_scans`: App-only. Read persisted plugin-owned scan summaries for native Security navigation.
- `list_codex_security_global_findings`: App-only. Read the latest plugin-owned finding occurrence for each stable repository target and finding identity.
- `list_codex_security_repositories`: App-only. Read plugin-owned repository summaries and their latest scan state.
- `get_codex_security_scan_context`: Load the authoritative target, mode, optional user context, artifact directory, live progress, and optional selected finding for a launched scan. For a remedia…
- `rename_codex_security_scan`: App-only. Change the display name of a saved scan.
- `update_codex_security_scan_context`: Replace the context for a running scan. The next phase uses the new value; workers in the current phase keep their original context.
- `update_codex_security_scan_context_from_app`: App-only. Replace the context for the running scan attached to this workspace.
- `update_codex_security_scan_progress`: Record a meaningful live scan phase or coverage milestone in the Codex Security workbench.
- `complete_codex_security_scan`: Finalization only: validate and seal already-authored scan-manifest.json and findings.json, generate report.md, index findings, and mark the scan complete. For…
- `finalize_codex_security_scan_usage`: App-only. Record final measured usage after the scan
- `fail_codex_security_scan`: Permanently mark a launched Codex Security scan as failed only after a confirmed unrecoverable blocker. For explicit user cancellation, use cancel_codex_securi…
- `set_codex_security_finding_triage`: App-only. Persist a completed finding
- `request_codex_security_finding_remediation`: App-only. Queue a completed finding for Codex remediation before sending the host a generate or regenerate request.
- `request_codex_security_finding_remediation_action`: App-only. Durably claim an apply or verify handoff before asking Codex to perform the local working-tree operation.
- `set_codex_security_finding_remediation`: Persist the bounded local remediation workflow state for a completed finding. The UI may mark a request as queued; Codex records generated, applied, verifying,…
- `export_codex_security_findings`: App-only. Export retained local findings as JSON, SARIF, or CSV, or the saved threat model as Markdown without running another analysis. Findings require compl…
- `list_codex_security_findings`: App-only. Load one bounded page of indexed findings for a completed local scan, migrating validated legacy finding details when needed.
- `claim_codex_security_finding_remediation_resend`: App-only. Atomically take ownership of an unowned or stale remediation host request before resending it.
- `release_codex_security_finding_remediation_claim`: App-only. Release a locally owned remediation host request after message delivery fails.
- `cancel_codex_security_finding_remediation_request`: App-only. Roll back an owned remediation request after the user declines its host follow-up.
- `mark_codex_security_finding_remediation_delivered`: App-only. Seal host-message delivery ownership before Codex starts a remediation worker.
- `prepare_codex_security_review_items`: Generate the changed-file inventory for a diff scan.
- `list_codex_security_review_items`: Read one page of the diff or Deep scan discovery inventory.
- `record_codex_security_discovery_candidates`: Normalize and replace the selected diff scan
- `list_codex_security_candidates`: Read one page of diff or Deep scan discovery candidates.
- `record_codex_security_candidate_validations`: Record the diff or Deep scan candidate validation results.
- `record_candidate_attack_paths`: Record the diff or Deep scan candidate attack-path results.
- `record_codex_security_scan_draft`: Save the canonical threat model, findings and reviewState as an unsealed draft. Use complete:false as soon as a model is available, even with empty findings, t…
- `get_codex_security_completed_scan`: Read the selected scan
- `save_codex_security_artifact`: Save a supplemental document or evidence file with temporary or persistent storage. Use scanId for a running scan, or targetPath for standalone documents and s…
- `read_codex_security_artifact`: Read a saved supplemental artifact from temporary or persistent storage, including after an MCP restart. Use the same scanId or standalone targetPath, storage…
- `post_deep_scan_peer_note`: Share a lead or ask another discovery worker to check specific evidence. Pass its threadId to reply with your independent check. The host identifies the author.
- `read_deep_scan_peer_notes`: List this scan’s shared leads and questions, or pass a threadId to read that lead and its replies.
- `get_codex_security_finding_issues`: Validate a sealed scan and read its findings
- `record_codex_security_finding_issues`: Validate a sealed scan and save accepted provider creates or updates, or verified issue reuse, in the shared local issue history. Record the returned identity…
- `mark_codex_security_scan_handoff_delivered`: App-only. Record that the launched scan instructions were delivered to Codex.
- `claim_codex_security_scan_handoff_delivery`: App-only. Durably claim scan handoff delivery before sending continuation instructions to Codex.
- `release_codex_security_scan_handoff_delivery`: App-only. Release a failed scan handoff delivery claim so the app can retry it.
- `attach_codex_security_scan_continuation_thread`: App-only. Persist the normal local Codex thread created for a claimed scan handoff.

## Environment

- `CI_RESULTS`
- `SOURCE_REPOSITORY`
- `SOURCE_SHA`
- `SCAN_SOURCE`
- `GITHUB_REPOSITORY`
- `GITHUB_EVENT_NAME`
- `GITHUB_EVENT_PATH`
- `GITHUB_SHA`
- `PR_NUMBER`
- `GITHUB_OUTPUT`
- `GITHUB_STEP_SUMMARY`
- `GH_TOKEN` (secret)
- `RELEASE_PROGRESS_WATCH`
- `RELEASE_RUN_ID`
- `EVENT_NAME`
- `BASE_CHANGED`
- `SCANNER_SHA`
- `SCANNER_RUNTIME_LOCK`
- `SCANNER_TARBALL`
- `PORT`
- `TENABLE_ACCESS_KEY` (secret)
- `TENABLE_SECRET_KEY` (secret)
- `LD_LIBRARY_PATH`
- `RUNNER_TRACKING_ID`
- `CODEX_HOME`
- `RUNNER_ENVIRONMENT`
- `GITHUB_REF`
- `GITHUB_SERVER_URL`
- `GITHUB_WORKSPACE`
- `RUNNER_TEMP`
- `GITHUB_RUN_ID`
- `CODEX_SECURITY_STATE_DIR`
- `CODEX_SECURITY_PLUGIN_ROOT`
- `CODEX_SECURITY_PYTHON_COMMAND`
- `PYTHON`
- `XDG_CACHE_HOME`
- `CODEX_SECURITY_SCAN_ROOT`
- `CODEX_MCP_NODE_PATH`
- `CARGO_CFG_TARGET_OS`
- `CODEX_SECURITY_TEST_WINDOWS_SYMLINKS`
- `CODEX_SECURITY_DEEP_SCAN_CONFIG_PATH`
- `CODEX_SECURITY_STARTED_AT`