Changelog

All notable changes to this project are documented here. The format is loosely based on Keep a Changelog.

Unreleased

Added

Fixed

• Cursor plugin MCP metadata now launches the published local stdio package instead of routing users to the public Worker endpoint, which is not configured as a turnkey fresh-scan service.
• Windsurf guidance now recommends local stdio or a configured self-hosted Worker, and Cursor/Claude plugin manifests are synchronized to 0.3.7.

[0.3.31] - October 30, 2026

Fixed

• Response-text citation matching now counts a brand link appearing after the first 20 retained evidence hosts, while preserving the evidence cap.
• Provider-native brand citation matching now recognizes links beyond the first 30 retained evidence URLs across OpenAI, Grok, Perplexity, Google AI Mode, and Google AI Overviews; strict host matching and bounded evidence remain unchanged.

[0.3.30] - October 29, 2026

Added

• Added portable Agent Skill onboarding for compatible MCP hosts, documenting the existing local package and provider-key boundaries.

Fixed

• Gemini visibility parsing now combines usable non-thinking text parts from multipart responses instead of losing later response text.
• Anthropic visibility parsing now combines all user-visible response text blocks rather than discarding later blocks.

Security

• Added repository ignore coverage for root-level plaintext secret-copy files to reduce accidental source-control exposure; this does not change runtime credential handling.

[0.3.29] - October 26, 2026

Fixed

• OpenAI Batch result collection now rejects missing requested prompt results instead of silently returning incomplete audit evidence, while retaining unknown-prompt-ID validation precedence.
• Brand citation matching now handles URLs followed by sentence-final punctuation without treating punctuation as part of the URL.

[0.3.28] - October 25, 2026

Fixed

• OpenAI Batch jobs with terminal failed, expired, or cancelled statuses now close instead of looping indefinitely as in_progress. If a failed terminal job has output/error files, collect and retain its validated partial responses, explicitly report missing prompts, and preserve provider failure status.
• Clarified that local storage differs from optional provider-side API processing in user-facing privacy documentation.

Added

• Local MCP onboarding for Devin, Perplexity Portable, OpenHands CLI, and Kimi Code CLI, using the existing npm package.
• Offline regression tests for Batch JSONL submission, upload-error behavior, terminal failures, partial results, and cancelling-versus-cancelled handling.

[0.3.27] - October 24, 2026

Security

• Patched MCP SDK HTTP tool-input denial-of-service risk (GHSA-46jr-4wx8-v453 / CVE-2026-111768) by upgrading to compatible MCP SDK 1.32.1, server/client 2.3.1, and Cloudflare Agents 0.30.1, enforcing a 1,024-element tool-input limit, and bounding modern and legacy hosted request bodies to 256 KiB. Updated vulnerable transitive runtime dependencies and retained zero known production audit vulnerabilities.

Fixed

• Corrected Anthropic Haiku cost guidance to use token-based API pricing instead of universal per-prompt estimates or guaranteed free credits.

[0.3.26] - September 24, 2026

Added

• Added current local MCP onboarding for Google Antigravity, JetBrains Junie, GitHub Copilot cloud agent/code review, Qwen Code, and Augment Code / Auggie while reusing the existing published package and secret-safe provider boundaries.

Fixed

• Upgraded @cloudflare/workers-oauth-provider to 1.0.0 and bound OAuth grants/access tokens to the deployment's canonical ${SELF_URL}/mcp RFC 8707 resource, preserving self-hosting without hardcoding the DigestSEO production hostname.
• Self-hosted Worker setup now makes the post-first-deploy SELF_URL update mandatory before connecting an MCP client so OAuth resource identity cannot silently point at another deployment.

[0.3.25] - September 24, 2026

Added

• Added current local/self-hosted MCP onboarding for Langflow, Cherry Studio, Flowise, and Kilo Code while reusing the existing @digestseo/mcp-geo package and provider-secret boundaries.

Changed

• The reusable AI Visibility Audit report prompt now consumes the shipped top_sources citation summary before individual citation examples, without inventing citation-share or source-category metrics.

Fixed

• All six hosted visibility tools now explicitly publish readOnlyHint, destructiveHint, and openWorldHint, improving safety metadata compatibility for MCP clients and current public-plugin validation without changing tool behavior.

[0.3.24] - September 23, 2026

Added

• Added Replit Agent onboarding for a user's self-hosted remote Worker, using Replit's custom MCP flow and OAuth dynamic client registration without misrepresenting the public DigestSEO endpoint as a turnkey scan service.
• Added Jan Desktop / Jan Agent onboarding for the published local stdio package and a user's self-hosted remote OAuth endpoint.
• get_citations / visibility.citations now returns a bounded top_sources summary from engine-native citation URLs, including deterministic citation/prompt counts, contributing engines, a representative URL, and whether each source is the tracked brand domain.

Changed

• Hosted SDK-v2 discovery now exposes the human-readable GEO Tracker by DigestSEO title, concise product description, and canonical product URL while preserving stateless modern/legacy compatibility.

[0.3.23] - September 23, 2026

Added

• Added native AnythingLLM local stdio onboarding through MCP Management / anythingllm_mcp_servers.json, including zero-key discovery and secret-safe optional provider configuration.

Changed

• Grounded Grok visibility scans now use xAI's current grok-4.7 model through the existing Responses API + required Web Search contract.
• Perplexity setup guidance now describes the already-shipped Agent API fast preset and live usage-based pricing instead of stale Sonar/fixed per-prompt wording.

[0.3.22] - September 23, 2026

Fixed

• Perplexity visibility scans now use the Agent API fast preset instead of the retiring fixed Sonar model selector, preserving grounded web search before Sonar tiers retire on September 27, 2026.
• Portable Agent Plugin and Claude marketplace metadata now advertise Google AI Mode consistently with the seven-surface product.

Changed

• Added Amazon Q Developer CLI onboarding for the published local stdio package, including zero-key tool discovery and secret-safe provider configuration.
• Self-hosting guidance now matches the stateless SDK-v2 hosted MCP route and current manual-refresh behavior.

[0.3.21] - September 22, 2026

Changed

• OAuth-protected hosted MCP traffic now uses the stateless SDK-v2 createMcpHandler path with legacy stateless compatibility and explicit Host/Origin validation; the old Durable Object binding is retained only as a migration hold.
• All twelve MCP tool schemas now use lean example-led descriptions and complete input-parameter descriptions, with regression coverage to prevent context bloat and missing guidance from returning.

[0.3.20] - September 22, 2026

Fixed

• Provider HTTP calls now have a 90-second timeout, so a stalled upstream API cannot hang a local or hosted visibility scan indefinitely.
• Authenticated /admin/run-live scans can set wait_for_completion: true so manual/operator refreshes stay attached to the HTTP request until per-engine service-binding work finishes instead of depending on Cloudflare's post-response waitUntil() window.

[0.3.19] - September 22, 2026

Added

• Added opt-in Google AI Mode visibility scans through SerpAPI, with separate per-prompt requests and citation/source capture when returned.
• Added update_brand so local MCP users can correct tracked-brand identity, competitors, aliases, exclusions, and refresh cadence without replacing active prompts or historical runs.

Changed

• track_brand and update_brand now accept refresh_frequency: manual; self-hosted scheduled refresh skips manual brands while explicit refresh_brand remains available.

[0.3.18] - September 21, 2026

Added

• Added opt-in xAI Grok visibility scans using grok-4.6 with required Web Search grounding, citation capture, local and self-hosted Worker key plumbing, and synchronized MCP/client metadata.

[0.3.17] - September 21, 2026

Added

• Local MCP users can set an exact 1-50-question measurement set with set_prompts; the operation preserves historical runs, rejects duplicate/blank questions, and is a no-op when the active set is already identical.

[0.3.16] - September 21, 2026

Fixed

• Gemini now defaults to gemini-3.1-flash-lite, avoiding the Gemini 2.5 access restriction Google applies to some new projects while preserving the same GenerateContent API integration.

[0.3.7] - September 19, 2026

Fixed

• Prompt generation now persists exactly the requested number of unique prompts or leaves the existing prompt set untouched, protecting fixed-scope audit runs from silently undersized prompt sets.

[0.3.6] - September 18, 2026

Added

• Gemini CLI extension metadata plus direct one-click install guidance for VS Code and LM Studio.

Fixed

• Live ChatGPT scans use grounded web search, domain-lookalike mentions no longer inflate brand/competitor scoring, and the conflicting GitHub Sponsors rail was removed.
• OpenAI model/pricing documentation was refreshed against the live search-backed implementation.

[0.3.5] - September 17, 2026

Added

• Owned privacy-policy and directory-readiness disclosure for local/Desktop distribution.
• Transparent audit score-calculation documentation.

Fixed

• Perplexity visibility migrated to the current Agent API.
• Cursor/plugin installs use the published local stdio package instead of the unconfigured public Worker, with provider-key configuration preserved.
• MCPB bundles remain production-only.

[0.3.4] - September 15, 2026

Fixed

• Replaced the local/MCPB better-sqlite3 dependency with Node's built-in node:sqlite, eliminating third-party native .node binaries that Claude Desktop on macOS can reject under Library Validation.
• The one-click MCPB and local stdio runtime now require Node.js 22.13+, where node:sqlite is available without an enable flag.
• Refreshed the MCP SDK and security-sensitive transitive pins so the production dependency tree audits with zero known vulnerabilities at release qualification time.
• Official MCP Registry metadata no longer advertises the public geo-mcp.digestseo.com/mcp remote while that endpoint is not configured as a turnkey fresh-scan service; the npm stdio package remains the canonical registry package.

[0.3.3] - September 10, 2026

Added

• Optional EUR 99 one-time AI Visibility Audit for teams that want a client-ready baseline without operating mcp-geo themselves; the open-source package remains free.
• Direct prefilled audit-request links with deterministic source markers on the repository README, plus a repo-native announcement for the same bounded offer.

Changed

• Package, MCP Registry, MCPB, Cursor, Claude plugin, and Worker-reported version metadata are synchronized on 0.3.3.
• The public /audit implementation prefills the exact offer and buyer inputs; later sample-report and methodology polish is merged in the repository but still awaits production Worker promotion.

[0.3.2] — July 27, 2026

Published to npm as @digestseo/mcp-geo.

Added

• Hosted MCP tool metadata: hosted tools use visibility.* namespaces and provide typed input/output schemas; the local stdio CLI retains flat tool names.
• Dedicated Worker storage configuration: Worker examples and migration documentation target the dedicated mcp-geo-db D1 database.
• Marketplace metadata: Cursor and Claude Code plugin manifests, along with refreshed MCP Registry, MCPB, Cline, and Smithery metadata for the DigestSEO-branded server and hosted endpoint.

Changed

• Package, registry, MCPB, and Worker-reported metadata are synchronized on version 0.3.2 and the @digestseo/mcp-geo package identity.
• Production transitive dependencies are pinned to patched Hono, fast-uri, and body-parser versions.

[0.3.0] — July 2026

A second distribution path: a local stdio CLI where users bring their own API keys, published to npm as @digestseo/mcp-geo, plus registry/marketplace metadata. The Cloudflare Workers path keeps every 0.2.1 behavior — the accuracy and security fixes now live in the runtime-agnostic core shared by both.

Added

• Local stdio CLI (npx -y @digestseo/mcp-geo): the same MCP tools backed by a local SQLite database (~/.digestseo/digestseo.sqlite, override with DIGESTSEO_DB_PATH). Engines run inline and sequentially — no fan-out needed locally. All logging goes to stderr; stdout is the JSON-RPC channel.
• Local brand-management tools (CLI only): track_brand (creates a brand + generated prompt set; falls back to three starter prompts without ANTHROPIC_API_KEY; accepts the per-brand aliases and exclude_terms from 0.2.1), list_brands (brands with active prompt counts), generate_prompts (regenerate a brand's prompt set via Claude Haiku). On Workers deployments these operations remain behind the X-Seed-Secret-gated /admin/* routes; the Worker MCP surface still exposes exactly the original six tools.
• src/core/ — runtime-agnostic core shared by the Worker and the CLI: engine callers, scoring, prompt generation, content-gap analysis, brand seeding, and registerTools/registerLocalManagementTools.
• src/db/types.ts (Db contract), src/db/d1.ts (D1 adapter), src/db/sqlite.ts (better-sqlite3 adapter with a _migrations-tracked migration runner). Both adapters carry the 0.2.1 brand columns (aliases_json, exclude_terms_json) and the partial-run-aware visibility-history query.
• Packaging/metadata: npm bin + mcpName, server.json for the official MCP registry (npm package + templated Worker remote in one manifest), manifest.json + .mcpbignore for the MCPB desktop extension, Dockerfile, glama.json, llms-install.md, README Quick Install section, CI + release-publish GitHub Actions workflows.
• Unit tests for the sqlite adapter and core seeding, alongside the 0.2.1 matching/citations/scoring suites (npm run test:unit runs all of them), and a stdio smoke test that drives the built CLI over JSON-RPC (npm run test:stdio).

Fixed

• Fallback seed prompts no longer contain a literal $CATEGORY placeholder — it is substituted with the brand's category (or name) before insert. Affects the Worker's /admin/seed fallback path too.

Changed

• hashPrompt keeps 0.2.1's HASH_FIELD_SEP = '\x1f' field separator (cache keys are unchanged from 0.2.1).
• npm manifest: dependencies now lists only what the published CLI actually needs at runtime (@modelcontextprotocol/sdk, better-sqlite3, zod); Worker-only packages (@cloudflare/workers-oauth-provider, agents) are devDependencies — wrangler bundles them into the Worker regardless, and npx @digestseo/mcp-geo installs stay small. This supersedes 0.2.1's "manifest tells the truth" arrangement, which predated the npm package.

[0.2.1] — June 2026

A security + accuracy pass ahead of wider distribution.

Added

• Optional CONNECT_SECRET secret. When set, the browser step of the OAuth connect flow shows a one-field form and only completes when the secret matches. Without it the OSS build keeps its auto-completing single-dev-user flow — which means anyone who knows the worker URL can connect an MCP client and call refresh_brand (spending your engine API credits). Recommended for every deployment whose URL is shared anywhere.
• Per-brand aliases and exclude_terms (migrations/0005_brand_alias_exclude.sql, Brand in src/db.ts, SeedBrandInput in src/seed.ts). Aliases are extra terms that always count as a mention; exclude terms suppress the bare-word match on the brand name and domain root — so "Monday" the brand stops matching "monday" the weekday — while the full domain and aliases still count. extractCitations now matches through a mentionsTermSet term set; its exported signature is unchanged. Apply 0005 to add the aliases_json + exclude_terms_json columns; existing brands read as empty arrays and behave exactly as before.
• SECURITY.md — security model, trust boundaries, and private vulnerability reporting.
• GitHub Actions CI (.github/workflows/ci.yml): tsc --noEmit + unit tests on every push and PR.
• Unit test suite (npm run test:unit, Node test runner + tsx): tests/matching.test.ts covers brand/competitor mention matching (excluded bare words, full-domain matches that survive exclusion, root terms not matching inside a larger word); tests/unit/citations.test.ts and tests/unit/scoring.test.ts cover citation extraction (extractCitations, hostMatchesDomain) and score aggregation (computeOverallScore). These are the pure functions every score flows through; the smoke suite still covers the deployed Worker end-to-end.
• migrations/README.md — how to apply migrations and why files must never be renumbered.
• README: Claude Code connect instructions (claude mcp add --transport http), architecture diagram, security section.

Fixed

• Word-boundary brand/competitor matching in src/openai.ts. mentionsTerm used haystack.includes(root), a raw substring test: a brand whose domain root is a common word (monday.com → "monday", notion.so → "notion") false-positived on the everyday word, and any term matched inside a larger word ("motion" inside "promotional"). Matching now uses Unicode-aware boundary checks — (?<![\p{L}\p{N}])…(?![\p{L}\p{N}]) — for the brand name and the domain root, while the full domain is still accepted as a high-confidence substring. The exported extractCitations signature is unchanged. (The homograph case — a bare root term that is itself a common word, e.g. "monday"/"notion" — is handled via the new per-brand exclude_terms.)
• Linked-citation checks require the exact brand domain or a subdomain of it. Previously a substring check meant notacme.com counted as a link to acme.com. Applies to extractCitations, the Perplexity and AI Overviews engine-citation merge, and get_citations' cited_url selection — all through the new shared hostMatchesDomain helper.
• NUL → unit-separator hash delimiter in hashPrompt (src/openai.ts). The field join used two literal NUL (\x00) bytes, which made git classify the whole file as binary and refuse to render its diffs. It now joins with the ASCII unit separator through a named HASH_FIELD_SEP = '\x1f' constant. Hash outputs change, so existing shared_prompt_cache rows simply miss and expire on their TTL — no cache clear or migration required.
• get_visibility_history matches check_visibility's stance on partial runs. Runs interrupted mid-flight (stuck at in_progress) now count toward history via COALESCE(completed_at, started_at), and runs with zero ok rows are excluded entirely instead of charting as a fake score of 0.

Changed

• Docs (README.md, SETUP.md) now recommend OpenAI + Anthropic (Claude) as the starting engine pair instead of leading with Gemini's free tier. Gemini capacity varies by model, project, and usage tier, so project-specific 429s can exclude Gemini rows from scoring; Google AI Overviews also frequently returns NO_AI_OVERVIEW (scored as a zero). The cheapest documented path could therefore produce misleading first-run data. Gemini and SerpAPI stay documented as opt-in engines; engine availability is unchanged and remains key-driven (getAvailableEngines untouched).
• SEED_SECRET and CONNECT_SECRET comparisons are constant-time.
• MCP server version string now tracks the package version (was stuck at an older value).
• Runtime dependencies (@cloudflare/workers-oauth-provider, @modelcontextprotocol/sdk, agents, zod) moved from devDependencies to dependencies — wrangler bundles either way, but the manifest now tells the truth.

[0.2.0] — May 2026

A correctness + architecture pass driven by an end-to-end debugging session on the production fork. Every fix is described in terms of what the prior version got wrong.

Added

• migrations/0004_response_status.sql: status (ok / failed / skipped) and error_message columns on prompt_responses, plus an index on status. Failed engine calls used to write raw_response='ERROR: ...' rows that downstream scoring treated as real zero-mention hits. They're now explicit status='failed' rows that aggregates exclude.
• POST /admin/run-engine — single-engine handler. /admin/run-live self-fetches into it once per engine so each engine runs in its own worker invocation with its own free-plan 50-subrequest budget. Idempotent on (run_id, engine).
• POST /admin/cleanup-failed-runs — one-shot deletion of legacy polluted rows that were written before the status column existed. Idempotent, optional brand_id filter.
• bulkCacheGet helper in src/db.ts — one D1 read for N prompt hashes instead of N separate reads.
• persistEngineRun helper in src/db.ts — one batched D1 write at the end of an engine run instead of per-prompt inserts + cache puts + per-chunk run updates.
• EnginePromptResult interface — engines collect their per-prompt results in memory and flush in one batch.
• runEngineInProcess in src/engines.ts — the in-process dispatcher used by /admin/run-engine.
• Fetcher service binding (env.SELF) in Env and EnginesEnv, plus "services" block in wrangler.example.jsonc. The per-engine fan-out goes through env.SELF.fetch() because a public-URL fetch back to your own workers.dev hostname trips Cloudflare's "Worker called itself" guard (error 1042) and never lands.
• SELF_URL env var in wrangler.example.jsonc — the cron path needs to know the worker's canonical URL since it has no inbound request to derive an origin from.

Changed

• Each engine client (src/openai.ts, anthropic.ts, perplexity.ts, gemini.ts, ai-overviews.ts) now uses the bulk pattern: hash all prompts up front, bulkCacheGet once, run LLM fetches in CONCURRENCY=5 chunks, collect EnginePromptResult[], flush via persistEngineRun. Per-invocation subrequest count drops from ~89 to ~26.
• getLatestCompletedRun anchors on EXISTS(prompt_responses with status='ok') instead of runs.status='completed'. Partially-finished runs still surface their data instead of silently disappearing from MCP tool output.
• check_visibility, compare_competitors, and get_content_gaps use run.completed_at ?? run.started_at so they accept in-progress runs.
• get_visibility_history and get_citations SQL queries filter pr.status = 'ok' in the join/where clause so failed/skipped rows can't pollute aggregates.
• /admin/run-engine INSERT OR IGNOREs its runs row before calling the engine — D1's cross-region replication may not have caught up with /admin/run-live's INSERT INTO runs by the time the self-fetch lands, and a missing parent row would crash the engine's batched insert with FOREIGN KEY constraint failed.
• runEngines accepts an optional request argument and falls back to new URL(request.url).origin when SELF_URL isn't set, so HTTP-triggered code paths work even without the env var.

Fixed

• Subrequest-cap row loss: pre-fix, every engine deterministically wrote 10 rows out of 20 because chunk 3 hit the 50-subrequest cap. Now writes complete in one batch under the cap.
• Silent row drops from failed Gemini / AI Overviews calls: pre-fix the catch block's insertPromptResponse(failed) was itself a D1 subrequest that could fail, leaving zero rows for the prompt. The bulk batch holds failure results in memory and writes them in the same batch as the successes.
• Visibility tools dropping engines whose runs got stuck at in_progress: relaxed getLatestCompletedRun surfaces them.

[0.1.1] — May 2026

• Removed scripts/setup.sh — the bash script wasn't reliable across platforms (Git Bash on Windows in particular couldn't pipe stdin to wrangler's interactive prompts). Replaced with a fully manual, copy-pasteable walkthrough in SETUP.md as the canonical install path.
• SETUP.md is now self-contained — every wrangler command, every interactive prompt response, and the most common failure modes are documented inline.

[0.1.0] — May 2026

Initial public release.

Added

• 5-engine support: ChatGPT (gpt-4o-mini), Claude (claude-haiku-4-5), Perplexity (sonar), Gemini (gemini-2.5-flash-lite), and Google AI Overviews (via SerpAPI).
• 6 MCP tools: check_visibility, get_visibility_history, compare_competitors, get_citations, get_content_gaps, refresh_brand.
• Opt-in engine availability — engines are enabled only when their API key is present.
• D1-backed storage: brands, prompts, runs, prompt_responses, shared_prompt_cache.
• Cloudflare Cron Trigger (every 6h) that respects per-brand refresh_frequency (daily/weekly).
• Admin routes: /admin/seed, /admin/run-live, /admin/run-batch-submit, /admin/run-batch-collect, /admin/generate-prompts, /admin/trigger-cron-test. All gated by SEED_SECRET.
• Smoke test suite at tests/smoke.test.mjs (Node built-in test runner; no external deps).
• Claude-Haiku-powered prompt generator and content-gap analyzer with deterministic fallback recommendations.