Tessary
Open-source monitoring that finds why production AI agents fail
Tessary is a uptime & monitoring tool for Developers and Small businesses, offered as open source. Founded in in . Popular alternatives include Sentry, PagerDuty, Datadog.
About Tessary
Tessary is an open-source, Apache 2.0 licensed reliability platform for AI agents in production, built by co-founders Akhil Varma and Nidhi Nair. They found that grading each agent session cost about five times more than running it. Tessary watches every production trace instead of a sample, flags failures and behaviour changes, and explains their cause.
Agents send traces over OTLP or an SDK, normalised to the OpenTelemetry gen_ai conventions, with PII redaction before storage. Cheap classifiers check every trace for drift, tool errors, secret leaks, malformed output, frustration and groundedness. An LLM triage step groups real findings into cases, and root-cause analysis reads the failing traces and, if connected, your GitHub repo to name the likely cause. Self-hosting is free, Tessary Cloud is free up to 10,000 traces a month, and paid plans with custom classifiers are quoted per customer.
It suits teams already shipping agents with OpenTelemetry who need to catch rare failures that sampled evals miss. Datadog LLM Observability and Sentry add agent tracing to platforms many teams already run, while Braintrust and Langfuse focus on evals and tracing. Tessary is very new (its repository dates from August 2026) and paid prices are not public.
Tessary launched on Product Hunt on October 22, 2026, with the tagline "Stop your agents from failing silently in production"; no rank was available when checked. Buying tip: self-hosted instances send an anonymous heartbeat without trace content every six hours, which TESSARY_TELEMETRY_ENABLED=false turns off.
Key features of Tessary
- Classifiers run on every production trace with no sampling
- Built-in checks for drift, tool errors, secret leaks and malformed output
- LLM triage groups real deviations into trackable cases
- Root-cause analysis over failing traces and your GitHub repository
- OTLP ingestion over HTTP and gRPC with OpenTelemetry gen_ai conventions
- PII redaction before traces are stored
- One-command self-hosting with Docker Compose
- Custom classifiers written as plain-language failure descriptions (paid)
Pros and cons
Pros
- Apache 2.0 open source with free self-hosting
- Checks every trace instead of a sampled subset
- RCA links failures to specific commits when a repo is connected
- Free cloud tier needs no credit card
- Telemetry is documented and can be switched off
Cons
- Very new project with a short track record
- Paid plan prices are not published
- Frustration classifier requires your own OpenRouter or TypeSafe key
- Groundedness check needs an Apple silicon Mac or a GPU instance
- Hands off to a human and does not open the fix itself
Tessary pricing
Tessary uses a open source model.
| Plan | Price | Includes |
|---|---|---|
| Open Source | $0 | Self-host with Docker Compose; all core features |
| Cloud Free | $0 | 10k traces/mo, 1 GB, 30-day retention, $10 one-time RCA credit |
| Paid | Custom | Custom classifiers, managed dedicated instance, Slack support |
Source: tessary.ai · checked on Oct 22, 2026
How Tessary compares in Uptime & monitoring
- Rank by votes
- #1 / 17
- Starting price
- $0
- Free alternatives
- 4
0 votes from members
Category median is $24/mo
Free or open-source tools in Uptime & monitoring
Tessary ranks #1 of 17 uptime & monitoring tools by community votes and is not directly comparable on price.
Common use cases
- 1Catching a spike in tool-call errors after a prompt or code change
- 2Detecting secret leaks in agent outputs in production
- 3Tracing a support agent's bad answers back to a specific commit
- 4Monitoring user frustration trends across agent conversations
- 5Self-hosting agent monitoring where trace data cannot leave your cloud
Best for
Engineering teams running AI agents in production with OpenTelemetry traces who want every trace checked and failures explained.
Not ideal for
Teams that want a mature, fully supported observability suite or an eval-first workflow (Datadog, Braintrust).
Integrations
- OpenTelemetry (OTLP)
- OpenLLMetry
- OpenInference
- GitHub
- OpenRouter
- Docker
Our take
Tessary is a promising open-source way to monitor every agent trace cheaply and get root-cause analysis tied to your code. Try the free cloud or a self-hosted instance on real traffic, but treat it as an early-stage tool until it builds a longer track record.
FAQ
- What is Tessary?
- Tessary is a Uptime & monitoring tool: Open-source monitoring that finds why production AI agents fail.
- How much does Tessary cost?
- Tessary is Open source; there is no mandatory paid plan.
- What are the best alternatives to Tessary?
- Community-listed alternatives to Tessary include Sentry, PagerDuty, Datadog, Atlassian Statuspage.
- What does Tessary integrate with?
- Tessary integrates with OpenTelemetry (OTLP), OpenLLMetry, OpenInference, GitHub, OpenRouter, Docker and more.
- Which platforms does Tessary support?
- Tessary is available on web, linux.
- Who is Tessary for?
- Engineering teams running AI agents in production with OpenTelemetry traces who want every trace checked and failures explained.
Alternatives to Tessary
See all 10PagerDuty
▲ 0On-call scheduling, alerting and incident response for ops teams
Freemium · $25DevelopersEnterprise
More Uptime & monitoring tools
Reviews
No reviews yet. Be the first to share real experience.
Sign in to write a review.