Search and Fetch remain free within published rate limits even with a $0 Wallet; Agent and Browser are separate prepaid metered products. A new signup Wallet promotional grant is one-time, not recurring Search/Fetch credit. No card needed for the free Search/Fetch APIs.

The standalone Free plan has fixed search/record/recommendation/crawl limits, no card requirement, and can be used for commercial production sites; Grow and Grow Plus are distinct metered plans that can bill for overages. Confirm the plan before enabling paid usage.

Free is a distinct capped database tier. Adding a credit card upgrades the database to pay-as-you-go ($0.05 per 1.000 billed search/upsert requests), where the Free query cap no longer applies; no automatic unlimited free usage. Pro/Enterprise trials offered on request are separate from this ongoing Free plan.

Index capacity, storage, and availability depend on your deployment. Hosted Meilisearch plans and enterprise features are separate.

You supply compute and persistent storage and handle cluster operations and backups. Typesense Cloud is a separate managed paid service.

Cluster hardware, storage, security configuration, upgrades, and backups are your responsibility. Managed OpenSearch offerings incur separate costs.

Results depend on external engines, which may rate-limit or block instances. Server operation, network access, and privacy controls are your responsibility.

No standalone free plan. A payment card is required for identity verification (temporary $0/$1 authorization). Set prepaid balance to $0 and disable auto-reload / set usage limits to avoid charges. Unused free credits do not roll over. Attribution is required to receive free credits; storing API result data is prohibited without separate agreement. Commercial-use terms require review.

Credits are shared across API operations and consumption varies by endpoint/depth. Free requests stop when credits are exhausted until the monthly reset; higher usage requires paid billing. Commercial-use rights for the free tier require confirmation.

You provide and pay for compute, storage, backups, monitoring and operations. GPLv3-or-later obligations apply when redistributing covered software; managed cloud services and paid support are separate.

Free tier is rate-limited; cached, errored and failed searches are not charged. Commercial use of the Free plan is not confirmed; SerpApi historically described a non-commercial Free plan, so verify current terms before production use. Higher quotas require paid plans.

Operator supplies and pays for servers, object storage, backups and operations; officially supported on Linux; Quickwit targets log/observability search, not managed hosted search.

You provide compute, persistent storage, operations, backups and security; local Docker setup requires at least 4 GB RAM. Internal protocols are not authenticated by default: isolate hosts and configure access controls/TLS. Vespa Cloud is a separate managed service.

Sonic stores a term index and returns object identifiers, not full source documents; your application must retain source records separately. Operate and back up its RocksDB store yourself; no built-in clustering or automatic failover. Hosting costs are separate.

Self-hosted compute, disk, backup, access control and maintenance are your responsibility. The project documents no built-in high availability; Elasticsearch API compatibility is partial and not a drop-in Kibana replacement.

Self-hosting requires JVM resources, storage, network security, patching, backups and operational expertise; optional cloud infrastructure, managed services and third-party integrations cost extra.

You must run indexing during your site build and host the generated static search bundle yourself. Hosting/CDN bandwidth and CI build resources may cost money; large indexes increase build and client download costs. This is software, not a managed hosted search plan.

This is an embedded library, not the managed Orama Cloud plan. You manage indexes, persistence, deployment and compute; optional embeddings, AI models, hosting, or commercial cloud features may incur separate costs. See repository LICENSE.md for Apache-2.0 terms.

Tantivy is a Rust library rather than a managed or distributed search server; applications must provide their own API, index storage and hosting. Distributed search is outside its scope.

Embedded Go library, not a turnkey hosted search service; storage, compute, API and index lifecycle are the operator's responsibility. Some optional vector features depend on build configuration.

Kubernetes and AVX2-capable CPUs are required for the documented NGT agent. Infrastructure, object storage and operations are not free; Vald focuses on dense-vector similarity rather than keyword search.

Free cluster is distinct from temporary $100/30-day Serverless/Dedicated trial; intended for evaluation and small projects, with limited features and no production SLA. Commercial production suitability should be confirmed with provider.

Run and maintain your own Java 17+ server or Docker container, persistent index and storage; compute/bandwidth are not free. Secure the default admin credentials before exposing a node; GPL obligations apply when redistributing.

Requires a compatible OpenSearch backend and Java/Docker infrastructure; hosting, storage, indexing and operations cost extra. Community edition has no guaranteed support; paid commercial support is optional.

Lucene Core is an embeddable Java library, not a hosted search service or ready-to-run search server; applications provide APIs, infrastructure, indexing and operational support.

An embedded/local library, NOT a managed search cloud service; you supply indexing/storage/compute. GPL-2.0-or-later obligations may affect redistributed linked applications; check license compatibility before commercial distribution.

Client-side/in-process library, not a hosted search API or server. Supply your own dataset and application compute; performance depends on index size.

Local/embedded library, not a hosted search service. Index data must fit available application memory; CPU, storage, hosting and maintenance are your responsibility.

The free Apache-2.0 library runs in your own browser or Node.js application; server-side deployment and optional persistent database connectors incur separate infrastructure or database costs.

Self-hosted indexing, storage, CPU/RAM, access control and server maintenance are your responsibility; infrastructure and any Git hosting costs are separate. No managed hosted free tier is included.

Requires Bazel build or your own container deployment, separate indexing/search processes, and self-managed CPU, RAM, storage and repository access; hosting costs are not included. This is not a managed hosted search tier.

Pre-v1 project; breaking API changes are possible. Requires a SQLite environment that permits loading extensions; local compute/storage and any external embedding APIs are not included.

Client/server application must build and store its own search index; no hosted search infrastructure is included. Repository development is infrequent and large indexes use application memory.

Requires a compatible PostgreSQL deployment, your own compute/storage and operational maintenance. AGPL-3.0 source-sharing obligations can apply to modified network-served versions; Enterprise features, commercial licensing and ParadeDB Cloud are separate.

Build and host the search index and WASM assets yourself; website/CDN bandwidth and CI build resources are not included. It is not a managed hosted search API.

Free API key requires email; anonymous quota is IP-bound. Requests above free quota return HTTP 429 unless pay-as-you-go is explicitly enabled; paid overage costs $5 per 1.000 requests with a spend cap. Commercial terms not independently confirmed.

No vendor-hosted free storage or compute is included. Operate Docker/Kubernetes, disk, memory, replicas, backups and networking yourself; paid infrastructure and embedding/model services remain separate.

The free license does not provide hosted search, GPUs, compute, disk, backups or model inference. Docker/server mode requires operating your own infrastructure; embedded Python is a local option, not a vendor-hosted free cloud plan.

This is an embedded library, not free managed search hosting. Your application provides CPU, memory, disk, indexing, backups and any cloud infrastructure; separate embedding models or APIs may cost money. Observe Apache-2.0 notices.

NOT a free hosted search index or search API. Requires a separately operated Elasticsearch/OpenSearch cluster and app infrastructure; hosted clusters, compute, storage and traffic may cost money. The local Searchkit library itself is free for commercial use.

This is a local search library, not a hosted vector-search API or free cloud service. You supply memory, CPU, storage, embeddings, application hosting and backups; third-party embedding APIs and infrastructure can cost money. Observe Apache-2.0 license notices.

Vectorlite is a local SQLite extension, not free hosted search. Requires a compatible SQLite driver, supported binaries/build toolchain, storage, CPU and memory. Embedding generation, cloud hosting, backups and third-party APIs are separate costs. Follow Apache-2.0 terms.

Faiss is an embedded library, not a hosted vector database or free API. CPU/RAM/storage, optional GPU hardware and CUDA/ROCm dependencies, embedding generation and any cloud infrastructure are supplied separately and may cost money; follow MIT and third-party notices.

Requires your own server, storage, database and browser/crawler resources; no hosted search capacity is included. Respect target sites' crawling policies. Modified network deployments must comply with AGPL-3.0 source-sharing obligations.

This is not free hosted txtai.cloud or free LLM inference. You supply CPU/GPU, memory, storage, model downloads and API credentials; third-party model licenses, hosted inference, infrastructure and external APIs may have separate costs or commercial restrictions.