Orca LinkedIn Intelligence
Collect LinkedIn activity and autonomously turn it into structured, user-defined profile insights.
The README identifies the LinkedIn-data API, LLM provider, required keys, optional Supabase authentication, and the broad categories of data scraped, earning partial transparency and attribution credit. Deductions apply because it does not describe retention, logging, exact fields sent to the LLM, privacy controls, or key rotation; without Supabase configuration the app and API are open by default. The agent may initiate additional scraping without per-action confirmation, narrowed permissions, rate limits, or recovery controls. No red-line behavior is shown, but dependency security is supported only by version declarations and build CI, with no audit, frozen-install, or vulnerability-response evidence.
The README, package metadata, and CI broadly agree on Node 20, pnpm, Next.js, and the build path, while external-service prerequisites are stated. The README claims coverage for every scraper and the agent, but the supplied CI runs only the build—not tests or lint—and no test sources are included. The product depends on RapidAPI, an LLM service, and optionally Supabase without documented fallbacks, availability handling, or retry behavior. No source evidence demonstrates user-facing failure messages, so that criterion receives zero.
Sales, recruiting, and investing audiences are clearly identified, and the profile-URL plus requested-insights workflow is concrete. Runtime requirements, the main stack, provider choice, and optional authentication are also described. Deductions reflect undefined capability limits, supported data shapes, scraping depth, lawful-use constraints, and thresholds for autonomous follow-up scraping. Although the core is described as a reusable Node.js library, no library API or integration example is supplied.
The README has a coherent structure covering purpose, operation, stack, requirements, environment, installation, testing, contribution, and licensing; naming is mostly consistent across repository and package metadata. The complete MIT text justifies full license credit. Deductions apply for the absence of a real invocation example, sample structured output, FAQ, systematic limitations, and changelog. Version 0.1.0 is present without release or compatibility policy. Maintenance responsibility is only weakly indicated by the copyright holder and a reference to CONTRIBUTING.md, with no supplied maintenance channel or update commitment; the private application package and standalone-library positioning are also not reconciled.
Combining several kinds of LinkedIn activity into structured, streamed insights and collecting additional evidence when needed could provide meaningful value beyond manual browsing, supporting ordinary-use scores for usability and marginal value. Deductions apply because no output schema, complete result example, quality-control mechanism, or measured outcome is provided. Operation requires potentially metered scraping and LLM services, but there are no call limits, caching controls, budgets, latency expectations, or cost estimates.
The described workflow has some corroboration in package scripts naming scraper and agent test suites, while CI confirms a build path. However, core effectiveness, comprehensive test coverage, and scalability claims are not traceable to supplied implementation, results, or metrics. The material does not show source citations in generated findings, preservation of underlying LinkedIn evidence, uncertainty labels, or an explicit separation between observed facts, model inference, and evaluative judgment.
- Without Supabase configuration, every page and API endpoint is open by default; production deployment should add mandatory authentication, access controls, and key protection.
- The tool processes profiles, posts, comments, reactions, and relationship networks. Users should first verify that LinkedIn rules, provider terms, and applicable privacy and employment laws permit the intended use.
- The agent can initiate more scraping autonomously, but the supplied material shows no confirmation gate, collection cap, rate limit, budget control, or stop mechanism.
- Structured insights may be model inferences. Sales, recruiting, or investment use should require citations to underlying evidence, uncertainty labels, and human review.
- The supplied CI and files do not adequately substantiate test coverage, failure handling, or dependency-security practices.
What does this agent do, and when should you use it?
Orca is an AI agent for deep LinkedIn profile analysis, with the desired insights defined by the user. It uses the Fresh LinkedIn Profile Data API to collect profiles, posts, comments, reactions, top-post engagement, and interaction-network data. The analysis agent reasons over that evidence and can invoke additional scraping tools when a particular insight needs more information. Findings stream into a UI built with Next.js 16, TypeScript, and Tailwind CSS, covering areas such as pain points, current focus, values, expertise, network influence, communication style, and changing interests. Its core logic resides in `orca-ai/`, which is presented as a standalone library for integration into other Node.js projects and scaled execution.
A user supplies a LinkedIn profile URL and specifies the insights to extract. Orca gathers baseline profile, post, comment, reaction, and top-post engagement data, then reasons across that material and the associated interaction network to produce structured insights. When the available evidence is insufficient for a requested insight, the agent autonomously calls additional scraping tools. Results stream back to the UI while processing continues. The implementation uses LangChain with support for OpenAI, Anthropic, and other model providers, while the reusable analysis logic lives under orca-ai/.
- A salesperson researching a prospect before outreach can identify priorities, pain points, and current areas of attention from public activity.
- A recruiter evaluating a candidate beyond a résumé can examine expressed values, interests, communication style, and demonstrated expertise.
- An investor assessing a founder can map the founder's thinking, changing interests, and statements about positioning.
- A researcher with a specific question about a LinkedIn user can request structured findings about influence, interaction networks, or communication patterns.
- A Node.js team running profile analysis in volume can integrate the standalone
orca-ai/library into an existing application.
What are this agent's strengths and limitations?
- Analysis extends beyond profile fields to posts, comments, reactions, top-post engagement, and interaction networks.
- Users define the desired insights, and the agent can collect additional evidence autonomously when necessary.
- The documented outputs include concrete dimensions such as pain points, values, expertise, network influence, communication style, and interest changes.
orca-ai/separates the core logic as a standalone library suitable for Node.js integration and scaled execution.- LangChain support for OpenAI, Anthropic, and other providers reduces dependence on a single model vendor.
- Operation depends on both a Fresh LinkedIn Profile Data RapidAPI key and an LLM-provider key, introducing external-service dependencies and potential usage costs.
- The application requires Node.js 20+, pnpm, network access, and local environment configuration rather than offering an offline or zero-setup mode.
- No import example, function signature, or batch-integration recipe is provided for the standalone
orca-ai/library. - Authentication relies on optional Supabase configuration; without those credentials, every page and the API run without login protection.
- The supplied material does not document behavior for inaccessible profiles, missing data, scraper failures, or API rate limits.
How do you install or deploy this agent?
Install Node.js 20+ and pnpm, then run:
git clone https://github.com/dimimikadze/orca.git
cd orca
pnpm installCreate .env.local in the project root with at least:
RAPIDAPI_KEY=your_key
OPENAI_API_KEY=your_keyRAPIDAPI_KEY is for the Fresh LinkedIn Profile Data API. An API key for the selected LLM provider is also required; OpenAI is the default. For optional login protection, additionally configure NEXT_PUBLIC_SUPABASE_URL and NEXT_PUBLIC_SUPABASE_ANON_KEY.
How do you use this agent?
Run pnpm dev from the repository and open http://localhost:3000. In the UI, provide a LinkedIn profile URL and define the insights you want Orca to extract. It will collect baseline profile and activity data, request more data through scraping tools when needed, and stream the resulting analysis to the interface. The README states that orca-ai/ can be embedded as a standalone Node.js library, but it does not provide a copyable import or library-level API invocation.
FAQ
Which external services or paid credentials are required?
Is Orca restricted to OpenAI models?
Does the application require users to log in?
Do its tests require live LinkedIn API calls?
USE_LIVE_DATA = true.Can the analysis engine be used without the included UI?
orca-ai/ is described as a standalone library for Node.js projects, although no library-level usage example is included in the supplied documentation.