Yes, it fits: the README provides both an MCP server for Claude Desktop and Claude Code and a keyless public jobs API.

• The “Beyond the catalogue” section explicitly lists an MCP server for Claude Desktop and Claude Code.
• The API section states that the catalogue is exposed through a public, keyless HTTP API and that GET https://freehire.me/api/v1/jobs requires no credential.
• Responses use a predictable data, meta, or error structure, and jobs and companies are addressed by public slugs, which simplifies client integration.
• Keyless access is not unlimited access. The project insights identify rate limits, caching, service stability, and data-use permissions as concerns. The README does not define the MCP tool list, quotas, or retry behavior.

Therefore, it is suitable as a Claude-facing data source and a quick integration target. It is not yet possible to confirm enterprise-grade quotas, private catalogue access, or an SLA from the provided documentation.

Yes, it covers the workflow from remote-job discovery through application tracking, but it does not replace verification on the original employer page.

• The README reports 3.3M+ unique live postings across 124 ATS platforms, 97 Telegram channels, and 320 sources, matching a cross-source search requirement.
• Search supports faceted full-text filtering by region, work mode, seniority, skills, and salary. These fields are derived from curated dictionaries rather than guessed values.
• “Beyond the catalogue” lists saved searches with email or Telegram digests, shared boards, an application board, recruiter-mail linking, an append-only event ledger, and reminders.
• The project insights note that job closure uses soft-close and periodic crawling, so status can lag. Salary, skills, work mode, and seniority fields can also be missing; the original company page remains necessary for confirmation.

It is therefore suitable as a high-volume job-search workspace and automation source. The README does not promise that every remote listing is real-time, complete, or continuously valid.

It depends: freehire provides a complete self-hostable stack, but operating a 3.3M+ posting catalogue is not a lightweight single-process deployment.

• The Quick start section provides make up, which starts the api, web, postgres, meilisearch, redis, and s3 services for an end-to-end local setup.
• The Stack section assigns different responsibilities to PostgreSQL/pgvector, Meilisearch, Redis, and S3-compatible storage, so the deployment carries several stateful services to maintain.
• The README reports 3.3M+ unique live postings, 470,000+ companies, and 320 sources; full indexing and ingestion can therefore require substantial storage and compute.
• Migration changes are not reapplied to an existing PostgreSQL volume. Recreating volumes with docker compose down -v can remove existing data.

This is a reasonable foundation for an operator comfortable with containers, databases, and search indexes. It is probably excessive if the requirement is only a small, read-only job lookup service.

Yes, architecturally it fits: freehire is designed around OpenAI-compatible endpoints, provider fallback, and LLM observability, although model quality and cost still depend on your endpoint configuration.

• The Stack section says langchaingo accesses LLMs through any OpenAI-compatible endpoint, avoiding a hard dependency on one vendor.
• Bifrost provides per-user virtual keys, cross-provider fallback, and spend tracking, which supports multi-provider routing and budget control.
• Langfuse provides LLM call tracing and observability. The feature reference also lists AI fit analysis, CV tailoring, and interview rehearsal.
• OAuth and LLM credentials are optional. The Local development section says that features gated by unset credentials remain disabled.

The architecture therefore matches your integration requirements. However, the README does not name supported models, token costs per feature, latency expectations, quality thresholds, or fallback triggers. It cannot establish that a particular model will produce acceptable CV content.

Yes, it fits: the README defines source ingestion, normalization, and SQL-generated type-safe access as explicit architectural boundaries that align with Go and sqlc without an ORM.

• The “Why freehire?” section lists Workday, Greenhouse, Lever, Ashby, and iCIMS, and says postings are normalized into one schema and deduplicated on a stable key.
• It states that adding a company is one row in the board catalog, suggesting that many source additions are configuration-driven rather than full-system rewrites.
• The Stack section uses Go with Fiber v2 for the HTTP server and sqlc for type-safe database access generated from SQL, explicitly noting “no ORM.”
• The project insights describe separate ingestion adapters, normalization, deduplication, and soft-close lifecycle handling. However, the supplied README excerpt does not include the adapter interface, field mapping rules, or test requirements.

The extension model is therefore a good fit conceptually. Before implementing a source, you still need the omitted Adding a source, Sources, and Workers sections to verify adapter availability and scheduling requirements.

It depends: the technology stack is a strong match, but both web and extension depend on a repository-local design-system symlink and are not fully independent frontend packages.

• The Frontend section explicitly uses SvelteKit 2, Svelte 5 runes, Tailwind 4, and adapter-node, matching your stated frontend stack.
• The web application is server-rendered, lives under web/, and consumes the backend through a same-origin API. A purely client-rendered application or a different routing model would need adaptation.
• The README specifically says the shared tokens and components come from the sibling design-system package, which must be installed before building web/ or extension/; it is symlinked rather than copied.
• The repository contains substantial TypeScript and Svelte code, but the README does not define the extension’s browser support matrix or release process.

It is therefore a good fit for a SvelteKit-oriented workflow. Compatibility is less certain if your CI disallows symlinks or requires a documented browser compatibility matrix.