Skip to content
SPECIFICTION
Get listed

Respan (formerly Keywords AI)

respan.ai

From $16.58/mo Free plan WebAPISelf-hosted
Visit respan.ai
Spec score 7.8 #23 of 31 in AI Agent Observability Tools

What it is

Respan, formerly Keywords AI, is an AI gateway with observability, evaluation, and security testing features. Its OpenAI-compatible gateway routes requests to more than 1,000 models through one endpoint, with fallbacks, load balancing, automatic retries, and response caching. The gateway may add 50 to 150 ms of latency. Respan records LLM calls, tool runs, retrievals, and agent turns as trace spans with input, output, latency, and cost. Evaluators can combine LLM judges, deterministic code checks, and human review, including for sampled production traffic. Monitoring tracks requests, errors, cost, latency, and tokens, with threshold alerts through Slack, email, or webhook. The integrations directory lists 171 connections, including OpenTelemetry, LangChain, OpenAI, Anthropic, AWS Bedrock, Slack, and PostHog. Platforms are API, self-hosted, and web. Free is free; Team is $199/mo billed annually. Enterprise pricing is custom. The homepage states support for SOC 2, GDPR, and HIPAA, and says a BAA is available for healthcare teams.

Spec sheet

Product
Respan describes itself as an AI gateway with built-in observability, evaluations, and security testing.Read 4 Oct 2026 · respan.ai
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.Read 4 Oct 2026 · respan.ai
Integrations
The integrations directory lists 171 integrations, including OpenTelemetry, LangChain, OpenAI, Anthropic, AWS Bedrock, Slack, and PostHog.Read 4 Oct 2026 · respan.ai
Support
The pricing page lists Community Discord for Free, email support for Team, and a dedicated support engineer for Enterprise.Read 4 Oct 2026 · respan.ai
Company background
The About page says Respan, formerly Keywords AI, began at the University of Illinois and joined Y Combinator's Winter 2024 batch.Read 4 Oct 2026 · respan.ai
Enterprise security
The Enterprise plan lists self-hosted deployment, Google and SAML SSO, HIPAA compliance, and a HIPAA BAA.Read 4 Oct 2026 · respan.ai
Evaluations
Evaluators can combine LLM judges, deterministic code checks, and human review, and can score sampled production traffic.Read 4 Oct 2026 · respan.ai
Gateway
Its OpenAI-compatible gateway routes requests to 1,000+ models through one endpoint.Read 4 Oct 2026 · respan.ai
Gateway latency
The gateway quickstart says the gateway may add 50 to 150 ms and may not suit products with strict latency requirements.Read 4 Oct 2026 · respan.ai
Monitoring
The product tracks requests, errors, cost, latency, and tokens, with threshold alerts sent through Slack, email, or webhook.Read 4 Oct 2026 · respan.ai
Privacy and compliance
The homepage states Respan supports SOC 2, GDPR, and HIPAA, and says a BAA is available for healthcare teams.Read 4 Oct 2026 · respan.ai
Reliability
The gateway provides model fallbacks, load balancing, automatic retries, and response caching.Read 4 Oct 2026 · respan.ai
Tracing
Respan captures LLM calls, tool runs, retrievals, and agent turns as spans in a trace with input, output, latency, and cost.Read 4 Oct 2026 · respan.ai

Plans and prices

Free Free 100k logs · 1k scores · 5 datasets · 2 evaluators · 5 prompts Read 4 Oct 2026 · respan.ai
Team $199 a year Billed yearly 10k scores · 100k logs · unlimited datasets, evaluators, and prompts · 5 members Read 4 Oct 2026 · respan.ai
Enterprise Priced on request Custom packages · volume discount · custom SLAs · dedicated support engineer · HIPAA BAA Read 4 Oct 2026 · respan.ai

Every Respan (formerly Keywords AI) plan and price →

Strong points and limits

Strong points

  • Routes requests to 1,000+ models through one endpoint.
  • Traces calls, tool runs, retrievals, and agent turns.
  • Evaluations can combine judges, code checks, and human review.
  • Threshold alerts cover requests, errors, cost, latency, and tokens.
  • API, self-hosted, and web platforms are listed.

Limits

  • Gateway may add 50 to 150 ms of latency.
  • Team costs $199/mo billed annually.
  • Enterprise pricing is custom.

Who it suits

Respan suits teams that need to route model requests while tracing and evaluating AI workflows in production. Its stated audience includes startups and enterprise teams.

Verdict

Respan combines model routing with tracing, evaluations, monitoring, and security testing. Its gateway’s stated 50 to 150 ms latency addition may matter for products with strict latency requirements.

Where Specifiction places it

Alternatives to Respan (formerly Keywords AI)All alternatives →