Traversal

A sales-led "AI SRE" that investigates alerts and incidents across the monitoring tools you already run.

ProprietaryPaid#ai-sre#incident-response#root-cause-analysis#alerting

What is Traversal?

Traversal is a hosted "AI SRE": an agent that investigates alerts and incidents by querying the telemetry and change data you already have. You connect read-only integrations (Datadog, Grafana, Prometheus, Splunk, Sentry, CloudWatch, GitHub, PagerDuty and more), then ask questions in its web app, Slack, Teams, over MCP or through an API. Each answer is a root-cause analysis with a confidence level, a timeline and cited evidence. Workers are long-running agents that join incident or alert channels and start investigating on their own.

Traversal, a venture-backed New York company, sells only through its sales team: no public prices, and the website offers demos and a pricing call rather than a signup. It runs as shared SaaS, single-tenant SaaS, or a managed deployment in your own AWS account; an Apache-2.0 Connector reaches private data sources.

On 3 November 2026 it announced a $120M Series B at a $1 billion valuation. Workers have been generally available since 2 September 2026; Alert Workers are in public beta. It sits on top of Datadog, Grafana or Sentry rather than replacing them, and competes with the AI investigation features those vendors now build in.

Our review

3.0

A provisional score for a product we couldn't try. You can't evaluate Traversal without talking to sales: there are no public prices and no self-serve signup, and the results we could find (RCA accuracy, hours saved) come from Traversal's own case studies. The design is sensible for a 3am investigation: read-only by default, cited evidence and a timeline rather than a bare answer, and it works across several monitoring tools instead of locking you into one. Teams already on one vendor should first test that vendor's built-in AI investigation feature, which costs no new procurement.

Pros

+ Read-only by default and installs nothing on your hosts; an optional open-source Connector handles private networks
+ Works across 30+ sources, from Datadog and the Grafana stack to Splunk, Sentry, GitHub and PagerDuty
+ Answers come with a confidence level, timeline and cited evidence you can click through to the source

Cons

− No public pricing or self-serve trial; you can't try it without a sales conversation
− No independent benchmarks or community discussion yet; accuracy figures are vendor case studies
− Credit metering covers alert-triggered and Worker investigations too, so cost depends on your alert volume

Reviewed by SeekDeploy Editorial · updated 2026-11-15. Reviews are independent — we don’t accept payment for coverage or ratings.

Should you use Traversal?

A good fit if you're…

Large engineering organisations whose telemetry is split across several vendors (say Datadog for APM, Splunk for logs, Prometheus for metrics) and whose incidents cross those boundaries
On-call teams with a procurement process who want an agent to draft the first root-cause hypothesis in the incident channel
Regulated companies that need the AI to run inside their own AWS account with their own Bedrock models

Probably not for you if you're…

Small teams that want to try something this week: there is no self-serve signup or public price
Teams with no decent logs, metrics or traces yet; Traversal reads telemetry, it doesn't collect it
Teams fully on one vendor that haven't yet tried that vendor's built-in AI investigation

Pick something else if…

you're already on Datadog and want its built-in Bits Investigation agent without a second vendor → Datadog
you run Grafana Cloud and want Assistant Investigations over the same data sources and RBAC → Grafana
your incidents are mostly application errors; its Seer agent works from issues, traces and profiles → Sentry

Gotchas to know before you commit

You can't try it without talking to sales. The website offers a demo and a pricing call, not a signup, and there's no public price list.
The confidence label measures how complete the data was, not how likely the answer is to be right; Traversal's own docs say to keep a person in the loop for high-impact decisions.
Read-only is a default, not a guarantee: a token with write scope, a GitHub or GitLab connection marked write, or an MCP tool that changes things gives it write access. Check every credential's scope.
Automatic investigations from alerts, schedules and Workers use credits too. Usage limits are per organisation, member and Worker, and the docs say they may need your Traversal representative to switch them on.
Public evidence is thin: accuracy and time-saved figures come from Traversal case studies, and Hacker News had no discussion of it in the 90 days to 15 November 2026. Ask for a pilot on your own past incidents.

Traversal pricing

Checked Nov 15, 2026 against the official pricing page

Sales-led with no public prices. Every new investigation is metered in credits, including ones started automatically by alerts, schedules, service accounts and Workers. Traversal quotes on alert and incident volume, team size and deployment option, so the expensive part is likely to be noisy alert channels with Workers attached.

SaaS

Contact sales

Teams that accept a multi-tenant cloud service

Web app, Slack, Teams, MCP and API
Optional Connector and Processor in your network
SOC 2 Type II

Limit: Investigations metered in credits

Single-tenant SaaS

Contact sales

Organisations needing an isolated deployment

Dedicated deployment run by Traversal in its cloud
AWS PrivateLink option

Bring Your Own Cloud

Contact sales

Regulated teams keeping data in their own AWS account

Managed by Traversal in a dedicated AWS account you own
Models via providers, a private gateway or your own Bedrock

Limit: Needs a dedicated /20 VPC across three availability zones
Limit: Web app and sign-in still run outside the deployment

What it costs in practice

Pilot with one on-call team	Contact sales No public price. Traversal's pricing page says the quote depends on alert and incident volume, stack and team size; ask for the credit cost per investigation and how follow-ups are counted (the docs say they aren't metered).
Alert Workers on busy alert channels	Varies Each alert a Worker triages starts a metered investigation, so cost scales with alert volume. Set organisation and per-Worker limits in Settings → Usage before switching them on.

Estimates in USD from list prices. Your bill depends on usage, region, and discounts.

Health & activity

Updated Nov 15, 2026 · last 90 days

Closed sourceThe product itself isn't public. The figures below are for its public companion repo, so treat them as a partial view.

GitHub stars

0 forks

Commits

9 committers

Merged pull requests

Last push

Nov 13, 2026

Open issues & PRs

From InteractionLabs/traversal-connector (Traversal Connector, the open-source (Apache-2.0) relay you deploy in your network). Counts cover the last 90 days unless noted.