Skip to the content
Know your tech before you buy it. Makers: get listed from $49/mo

Respan (formerly Keywords AI)

respan.ai

  • Free plan
  • Web
  • API
  • Self-hosted
7.4 Club score Visit respan.ai

What Respan (formerly Keywords AI) is

Respan, formerly Keywords AI, is an AI gateway and observability platform for teams building and operating AI applications. Its OpenAI-compatible gateway routes requests to 1,000+ models through one endpoint, with fallbacks, load balancing, retries, and response caching. Respan can record model calls, tool runs, retrievals, and agent turns as trace spans, including input, output, latency, and cost. Teams can evaluate sampled production traffic using LLM judges, code checks, and human review. Monitoring covers requests, errors, cost, latency, and tokens, with threshold alerts through Slack, email, or webhooks. The integrations directory lists 171 options, including OpenTelemetry, LangChain, OpenAI, Anthropic, AWS Bedrock, Slack, and PostHog. Plans include a free tier and Team at $199/mo billed annually; Enterprise pricing is custom. Respan supports API, self-hosted, and web deployment. Its homepage states support for SOC 2, GDPR, and HIPAA, and says a BAA is available for healthcare teams. The gateway may add 50 to 150 ms of latency, which could be unsuitable for products with strict latency requirements.

Who it suits

Respan is aimed at startups and enterprise teams that need to trace workflows, evaluate agent behavior, and monitor quality in production. Its self-hosted option and listed enterprise security features may matter to teams with deployment or compliance requirements.

Good to know, and what to watch

  • Routes requests to 1,000+ models through one endpoint.
  • Traces include input, output, latency, and cost.
  • Evaluations can combine LLM judges, code checks, and human review.
  • Alerts can go through Slack, email, or webhook.
  • Free plan includes 100k logs and 1k scores.
  • Gateway may add 50 to 150 ms latency.
  • Team is $199/mo billed annually.
  • Free plan allows five datasets and two evaluators.

Enterprise

—

Free

Free

Team

$199/yr

Free plan

Getting started

  1. Open the Respan website
  2. Choose API, self-hosted, or web access
  3. Start with the Free plan or begin the available free trial
  4. Connect models and services through the OpenAI-compatible gateway
  5. Configure traces, evaluations, monitoring, and integrations

Limits to check

The Free plan includes 100k logs, 1k scores, 5 datasets, 2 evaluators, and 5 prompts. Team costs $199/mo billed annually and includes 10k scores, 100k logs, unlimited datasets, evaluators, and prompts, plus five members.

The Club's review

Respan, formerly Keywords AI, is an AI gateway and production observability suite for teams running language-model and agent workflows. It is most useful when routing, tracing, evaluation, and operational monitoring belong in one stack; teams with hard response-time targets should consider the gateway’s added delay before committing.

Overview

Respan provides an OpenAI-compatible endpoint that can route requests across more than 1,000 models. Fallbacks, load balancing, automatic retries, and response caching reduce the amount of reliability logic a team must build around individual providers. That consolidation is valuable for multi-model applications, but the gateway quickstart warns of an additional 50 to 150 ms, making it a questionable choice for latency-critical user experiences.

Its observability model follows a workflow rather than a single completion. LLM calls, tool runs, retrievals, and agent turns appear as spans containing input, output, latency, and cost. Evaluations can mix LLM judges, deterministic code checks, and human review, including scores from sampled production traffic. Monitoring covers requests, errors, cost, latency, and tokens, with threshold alerts delivered through Slack, email, or webhook.

Key features

Gateway reliability

The single endpoint and compatibility with OpenAI-style clients simplify provider changes. Fallbacks and retries help recover from model failures, while balancing and caching can smooth traffic and repeated requests. These controls suit teams operating several providers; a small application using one model may gain little from the extra gateway layer.

Tracing and evaluation

Span-level coverage of retrieval and tools is particularly relevant to agents, where a final answer can hide the step that failed. Combining automated judges, code assertions, and human review supports both repeatable regression checks and nuanced quality assessment. Production-traffic sampling links evaluation to behavior users actually encounter rather than only to prepared test sets.

Monitoring, integrations, and security

Cost and token tracking sit alongside latency and error metrics, so teams can investigate quality and spending in the same operational view. The integrations directory includes 171 entries, among them OpenTelemetry, LangChain, OpenAI, Anthropic, AWS Bedrock, Slack, and PostHog. Respan states support for SOC 2, GDPR, and HIPAA; healthcare teams can obtain a BAA, and Enterprise adds self-hosting, Google and SAML SSO, HIPAA compliance, and a HIPAA BAA.

Pricing

Respan uses a freemium model with a free plan and a free trial.

PlanPrice and limitsSupport and fit
FreeFree; 100k logs, 1k scores, 5 datasets, 2 evaluators, and 5 promptsCommunity Discord. Suitable for evaluating the platform or a small workflow, but the evaluator, dataset, and prompt caps constrain broader programs.
Team$199 a year, billed yearly; 10k scores, 100k logs, unlimited datasets, evaluators, and prompts, and 5 membersEmail support. The higher evaluation allowance and unlimited evaluation assets suit a small production team, while five-member access limits larger groups.
EnterpriseCustom packages, volume discount, custom SLAsDedicated support engineer, self-hosted deployment, Google and SAML SSO, HIPAA compliance, and HIPAA BAA. Intended for organizations requiring contractual support and controlled deployment.

The published paid entry is $199/mo when billed annually. Readers should distinguish that marketing entry from the Team table’s stated $199 a year billing term and confirm the applicable commercial terms before purchase.

Platforms

Respan is offered through an API, web interface, and self-hosted deployment. The combination supports hosted adoption while leaving Enterprise customers a self-managed option.

Who it's for

Startups and enterprise teams building LLM or agent workflows are the clearest audience. Choose Respan when one product must handle provider routing, multi-step tracing, production scoring, alerts, and spend visibility. It is less suitable when every millisecond matters or when you only need a basic model endpoint without evaluation and monitoring.

Pros and cons

  • Pros: Routes across 1,000+ models with fallbacks, balancing, retries, and caching; traces calls, tools, retrievals, and agent turns; combines LLM, code, and human evaluation; tracks cost and tokens with configurable alerts; offers broad integrations and Enterprise healthcare and SSO controls.
  • Cons: The gateway may add 50–150 ms; Free limits evaluators, datasets, prompts, and scores; Team is capped at five members; Enterprise pricing is custom.

Alternatives

For a broader directory of comparable products, see AI LLM Observability Tools, LLM Observability Tools, and AI Agent Observability Tools.

  • W&B Weave is a better fit when its Free plan’s 1 GB monthly ingestion and 5 GB storage, plus evaluations, tracing, and scorers, meet your needs.
  • Opik suits teams wanting open-source observability and evaluation code they can run locally.
  • Helicone is an open-source alternative with a free Hobby allowance of 10,000 requests and 1 GB storage.
  • Maxim AI fits a small developer workspace offering up to three seats and 10,000 logs per month on its free plan.
  • OpenLIT is preferable for unlimited self-hosted usage, users, projects, and environments at no charge.
  • Agenta offers a free Hobby tier with two team members, 5,000 agent runs monthly, and 20 evaluations monthly.
  • Portkey provides a free Developer tier with 10,000 recorded logs monthly and short retention windows.
  • Fiddler AI is an alternative for teams prioritizing real-time guardrails and latency below 80 ms.

Verdict

Choose Respan if your team needs a single control point for multi-model routing and detailed, production-grade tracing, evaluation, monitoring, and compliance workflows. Look elsewhere if the gateway’s 50–150 ms overhead is unacceptable, or if its free and five-member Team limits do not match your scale.

The verdict

Respan combines routing, tracing, evaluations, and monitoring for AI applications. Teams with strict latency requirements should account for the gateway's potential added delay.

Questions people ask

Is there a free plan?

Yes. Free is listed as free and includes 100k logs, 1k scores, 5 datasets, 2 evaluators, and 5 prompts.

What does Team cost?

Team is listed at $199/mo, billed annually.

Which platforms does Respan support?

It supports API, self-hosted, and web platforms.

Does it trace agents and retrievals?

Yes. Respan lists agent tracing and retrieval tracing, and captures those activities in trace spans.

Does Respan support compliance requirements?

Its homepage states support for SOC 2, GDPR, and HIPAA, with a BAA available for healthcare teams.

What the maker publishes

Company background
The About page says Respan, formerly Keywords AI, began at the University of Illinois and joined Y Combinator's Winter 2024 batch. source
Enterprise security
The Enterprise plan lists self-hosted deployment, Google and SAML SSO, HIPAA compliance, and a HIPAA BAA. source
Evaluations
Evaluators can combine LLM judges, deterministic code checks, and human review, and can score sampled production traffic. source
Gateway
Its OpenAI-compatible gateway routes requests to 1,000+ models through one endpoint. source
Gateway latency
The gateway quickstart says the gateway may add 50 to 150 ms and may not suit products with strict latency requirements. source
Integrations
The integrations directory lists 171 integrations, including OpenTelemetry, LangChain, OpenAI, Anthropic, AWS Bedrock, Slack, and PostHog. source
Intended users
The About page says Respan helps teams trace workflows, evaluate agent behavior, and improve quality in production, and that startups and enterprise teams use it. source
Monitoring
The product tracks requests, errors, cost, latency, and tokens, with threshold alerts sent through Slack, email, or webhook. source
Privacy and compliance
The homepage states Respan supports SOC 2, GDPR, and HIPAA, and says a BAA is available for healthcare teams. source
Product
Respan describes itself as an AI gateway with built-in observability, evaluations, and security testing. source
Reliability
The gateway provides model fallbacks, load balancing, automatic retries, and response caching. source
Support
The pricing page lists Community Discord for Free, email support for Team, and a dedicated support engineer for Enterprise. source
Tracing
Respan captures LLM calls, tool runs, retrievals, and agent turns as spans in a trace with input, output, latency, and cost. source

Checked against the maker's pages on 4 Oct 2026.