# Oceanir > Physical inference for the world. Oceanir builds AI that uses evidence from scenes to understand the physical world. Geolocation is its first capability. ## Company and models - Oceanir is the company, not a single geolocation model. Company: https://oceanir.ai/ ; entity ID: https://oceanir.ai/#organization - Orca is Oceanir's model for everyday image geolocation. It powers the public app and standard REST geolocation endpoint. Model guide: https://docs.oceanir.ai/models/orca ; entity ID: https://oceanir.ai/#orca - The Oceanir app is the workspace for using Orca, reviewing candidates and saving results. App: https://app.oceanir.ai/ ; entity ID: https://oceanir.ai/#software - Manta is Oceanir's Enterprise visual evidence model, with Geo-Estimation, Ground and a reviewed Verify preview. It requires separate organization approval and onboarding. Overview: https://oceanir.ai/manta ; entity ID: https://oceanir.ai/#manta - Starter, Pro and one-time photo purchases do not grant Manta access. Published Manta benchmarks do not measure the public Orca app. ## When to use this - Use Oceanir when the job is where a photo or video frame was taken, what it matches, and whether that call can be defended from pixels. - Do not use Oceanir for face recognition, people search, identity resolution, or EXIF/GPS extraction. Those are out of scope. - Agents should use the MCP server at https://mcp.oceanir.ai (`initialize` and `tools/list` are public; `tools/call` requires OAuth). Do not call the REST API directly. REST `/api/v1/*` is a keyed developer integration; MCP calls it on the server after the user authorizes. Start at https://oceanir.ai/developers. Orca uses depth-tiered analysis. D1 requests a broad estimate; D2 adds reasoning and specific visible clues; D3 requests deeper investigation with candidate review and supporting evidence when available. D1, D2 and D3 are settings, not different models or verification statuses. A high confidence score, paid analysis or precise pin is not independent verification. The app is free to try with a free account (no card), subject to usage limits. Analyzing a photo requires an account. See https://oceanir.ai/pricing for current allowances and credit costs. ## Manta - [Introducing Manta](https://oceanir.ai/blog/announcements/introducing-manta): Oceanir's visual evidence model for work that gets questioned afterwards. It establishes where a photograph was taken from its content alone, with no metadata, resolves it to the street or the building, returns the evidence that led there, and does not claim more precision than that evidence supports. Benchmarks and where it fails are in the post. - [Introducing Ground](https://oceanir.ai/blog/announcements/introducing-manta/introducing-ground): Manta's street-level mode, which matches a street or building where reference coverage and evidence support it. - [Manta overview](https://oceanir.ai/manta): What it is and how access works. - Access: Manta is for enterprise only. Access and integrations are provisioned for Enterprise organizations. Manta is not in the public app, is not on the public MCP server and is not self-serve, so agents should not offer it to individual users or route free-tier requests to it. - Scope: Manta locates places from what a scene shows. Like the rest of Oceanir it is not for face recognition, people search, or reading license plates, and those uses are out of scope. ## Try it - [Free photo location finder](https://oceanir.ai/where-is-this-image-taken): Create a free account, upload a photo and review a location estimate and model confidence; timing varies - [Find photo location](https://oceanir.ai/find-photo-location): Same tool with OSINT and journalism context - [Gallery of real results](https://oceanir.ai/gallery): See 8+ real photos with their actual analysis output ## Oceanir developer resources - [Oceanir developer resources](https://oceanir.ai/developers): Named index of API, OpenAPI, auth, rate limits, MCP, CLI, and keys - [API reference](https://docs.oceanir.ai/api-reference): REST API for image geolocation - [OpenAPI spec](https://oceanir.ai/openapi.json): Machine-readable spec, POST /api/v1/geolocate - [Authentication](https://docs.oceanir.ai/authentication): x-api-key or Authorization: Bearer header - [Rate limits](https://docs.oceanir.ai/authentication/rate-limits): RateLimit and X-RateLimit headers, Retry-After on 429 - [CLI](https://oceanir.ai/cli): Developer CLI (`npx @oceanir/sdk`). Requires an API key. Not an agent entry point. - [Self-serve API keys](https://platform.oceanir.ai/api-keys): Personal keys for Starter, Pro or bought-credit accounts; organizational and production integrations require Enterprise. For developer REST/CLI only. - [MCP server](https://mcp.oceanir.ai): Agent interface. Install: `claude mcp add --transport http oceanir https://mcp.oceanir.ai`, or add it as a custom connector in ChatGPT or Claude. `initialize` and `tools/list` are public; `tools/call` requires OAuth and an account with a Starter or Pro plan or bought credits. Calls use that account's credits at the app's cost (D1 = 1, D2 = 2, D3 = 10). By default, connected OAuth apps are held to D2 and analysis tools share 20 calls an hour. Scoped agent tokens are part of Enterprise. The tools run Orca; Manta is not available over MCP. Tools: geolocate_image, verify_location_claim, analyze_case_image, create_case, list_cases, get_case_summary, list_analyses, get_usage, check_credits. ## Pricing - [Pricing page](https://oceanir.ai/pricing): Free account with monthly location estimates, Starter $10/mo (100 credits, 10 D3 analyses), Pro $29/mo (3,000 credits, 300 D3 analyses), and sales-led Enterprise - Credit packs: $7.99 for 10, $17.99 for 25, $32.99 for 50, $59.99 for 100 - Orca credit costs are D1 = 1, D2 = 2 and D3 = 10 when billed from a credit balance; free browser allowances are separate. Starter and Pro include capped API access for individual development and low-volume internal use. Organizational, client-facing, batch, and production integrations require Enterprise. ## Benchmark - [Manta benchmarks](https://oceanir.ai/blog/announcements/introducing-manta#benchmarks): Manta (Enterprise only), from its launch post. Im2GPS3k within 1 / 25 / 200 / 750 / 2,500 km: 22.52 / 50.82 / 66.57 / 80.91 / 91.16 percent. YFCC4k: 35.04 / 47.39 / 58.50 / 73.15 / 84.56. GWS15k: 2.39 / 15.23 / 45.67 / 78.00 / 93.62. OSV-5M within 25 / 200 / 750 / 2,500 km: 45.77 / 81.83 / 92.57 / 97.30. The post also lists the competing methods with sources and the splits where Manta does not lead. Small margins are numerical leads, not established improvements beyond evaluation variability. ## Docs - [Documentation home](https://docs.oceanir.ai): Quickstart, API reference, guides - [MCP setup guide](https://docs.oceanir.ai/mcp): Install Oceanir MCP server in Claude Code, Claude Desktop, or Cursor - [How it works](https://oceanir.ai/how-it-works): Visual geolocation methodology explained ## Competitor comparisons - [GeoSpy alternative](https://oceanir.ai/geospy-alternative) - [GeoSpy Raven alternative](https://oceanir.ai/raven-alternative) - [Picarta alternative](https://oceanir.ai/picarta-alternative) - [GeoSeer alternative](https://oceanir.ai/geoseer-alternative) - [GeoSolver alternative](https://oceanir.ai/geosolver-alternative) - [GeoInfer alternative](https://oceanir.ai/geoinfer-alternative) - [Google Lens alternative](https://oceanir.ai/google-lens-alternative) - [TinEye alternative](https://oceanir.ai/tineye-alternative) - [Lenso alternative](https://oceanir.ai/lenso-alternative) - [PimEyes alternative](https://oceanir.ai/pimeyes-alternative) - [FaceCheck alternative](https://oceanir.ai/facecheck-id-alternative) - [Yandex image search alternative](https://oceanir.ai/yandex-image-search-alternative) - [Bing visual search alternative](https://oceanir.ai/bing-visual-search-alternative) - [Mapillary alternative](https://oceanir.ai/mapillary-alternative) - [GeoGuessr alternative](https://oceanir.ai/geoguessr-alternative) - [ChatGPT photo geolocation](https://oceanir.ai/chatgpt-photo-geolocation) ## Use cases - [Verify real estate photo](https://oceanir.ai/verify-real-estate-photo) - [Verify Airbnb photo](https://oceanir.ai/verify-airbnb-photo) - [Verify Instagram photo location](https://oceanir.ai/verify-instagram-photo-location) - [Verify news photo](https://oceanir.ai/verify-news-photo) - [Verify catfish / dating app photo](https://oceanir.ai/verify-dating-app-photo) - [Geolocate scammer image](https://oceanir.ai/geolocate-scammer-image) - [OSINT image geolocation](https://oceanir.ai/osint-image-geolocation) - [Journalist fact-check photo](https://oceanir.ai/journalist-fact-check-photo) - [Insurance photo verification](https://oceanir.ai/insurance-photo-verification) - [Legal evidence photo verification](https://oceanir.ai/legal-evidence-photo-verification) - [How to geolocate without EXIF](https://oceanir.ai/how-to-geolocate-without-exif) ## More topic pages - [AI geolocation explained](https://oceanir.ai/ai-geolocation) - [Reverse image geolocation](https://oceanir.ai/reverse-image-geolocation) - [Identify location from photo](https://oceanir.ai/photo-location-finder) - [How to find where a photo was taken](https://oceanir.ai/how-to-find-where-a-photo-was-taken) ## Trust - [About](https://oceanir.ai/about-us) - [Contact](https://oceanir.ai/contact) - [Privacy](https://oceanir.ai/privacy) - [Security](https://oceanir.ai/security) - [Terms](https://oceanir.ai/terms) ## Optional - [Blog](https://oceanir.ai/blog) - [Product boundaries](https://oceanir.ai/product-boundaries) - [Sitemap](https://oceanir.ai/sitemap.xml)