The most complete eval and experiment workflow in its class, billed by gigabytes and scores (Starter free, Pro $249 a month); Starter with a card is often cheaper than Pro, stored provider keys were potentially exposed in a May 2026 incident (Braintrust says the attacker used only Anthropic keys), and BYOC, self-hosting, SAML and the BAA need Enterprise.
Free Apache-2.0 Python framework for explicit RAG and agent pipelines; the paid Haystack Enterprise Platform is quote-only and the library removes deprecated features within about a month.
Open-source LLM tracing and evals that is far cheaper than per-trace rivals (Core $29 a month, unlimited users), now owned by ClickHouse, with a forced v4 cutover on Langfuse Cloud on Nov 16 2026 and retention rules only from Pro.
Free MIT-licensed visual builder for agents, RAG and MCP servers with visible tool calls, but the hosted cloud is gone and an exposed server needs constant patching and lockdown.
A deep SEO data suite with plans from $129 to $1,499 a month and Brand Radar AI tracking from $50, but seats and add-ons stack up, rank tracking is weekly, accounts can be restricted automatically and Trustpilot shows 2.1 from 324 reviews.
A no-code AI agent builder with a free plan and credit plans from $9 to $999 a month: cheap for steady small-site volume, but a credit is a reply, agents go silent at zero, and independent evidence is one Shopify review.
A capable open-source AI app builder, free to self-host or $59 and $159 a month per workspace in the cloud, whose included credits buy only a few dollars of model use, under a modified Apache license and with 21 published security advisories.
Good official reference for Claude skills, but no tags or releases to pin and licences differ by folder.
Strong open-source RAG engine for hard documents with a cheap $29 cloud plan, but 1.0 is still a release candidate, defaults are weak and the cloud terms are unfinished.
A placeholder scaffold: its three scripts print Findings: 0 for any path, so do not install it expecting a working reviewer.
See the article verdict.
Fits chat, writing and coding work in OpenAI-style clients that want model choice on one key.
A 33-model catalog from 10 makers behind one endpoint; choose by context, listed rate and capability labels.
Connecting is a base URL and a key for any OpenAI-style client; the main things to understand are key scope, streaming, limits and 402.
A clearly priced OpenAI-compatible gateway to 33 models with token plans from free to $50.
Six clearly listed token plans from $0 to $50 with yearly billing at ten months' price; read the overage rates per model and the split 7-day/14-day refund windows before paying.
A free MIT-licensed agent framework with huge adoption, paired with LangSmith at $0 or $39 a seat whose trace overage rose from $0.50 to $5 per 1,000 in under a year; strong compliance, persistent developer criticism.
A very broad work platform with a free tier and seats from $7 to $12 a user a month billed yearly, but AI plans at $9 and $28 a user are extra, annual plans need 30 days notice to cut seats, and Trustpilot shows 4.2 from 716 reviews with recent billing complaints.
An Apache-2.0 AI memory layer with a free tier and hosted plans at $19 and $249 a month, but a steep ladder, vendor-run benchmarks, SOC 2 Type I only and mixed public user evidence.
A free, MIT-licensed AI workspace for desktop and Docker; the $50 and $99 cloud plans lack MCP and custom agents, GitHub lists 25 security advisories and the Android app is rated 2.7.