Tsuga
by Tsuga
Bring-your-own-cloud observability that keeps telemetry, and its cost, inside your own AWS account
Tsuga is an AI-native observability platform that deploys inside your own cloud account instead of shipping telemetry to a vendor's SaaS. Engineering and platform teams get logs, APM, infrastructure monitoring, observability pipelines and AI agent traces while data stays behind their own perimeter under customer-managed keys, and storage and compute land on their own cloud bill rather than a vendor markup.
Tsuga is a Paris-founded observability company built on a bring-your-own-cloud architecture: rather than ingesting telemetry into a vendor-operated cloud, the platform deploys into the customer's own account via infrastructure-as-code, a process the company says takes under two hours on AWS, with support also claimed for Microsoft Azure, Google Cloud and sovereign clouds. All data stays encrypted with customer-managed KMS keys in the customer's own object storage and, in the company's framing, cannot leave the cloud perimeter. The functional surface covers log management with field-level transformation and routing, application performance monitoring, infrastructure monitoring, observability pipelines and sensitive-data scanning; the AI-native part is that traditional application and infrastructure traces sit alongside AI agent traces, prompt and token visibility, confidence metrics and agent call graphs, which is the gap the company argues legacy tools have. The commercial model is the actual product thesis: a flat rate per gigabyte ingested with unlimited retention and unlimited fields, no markup on storage or compute because those costs hit the customer's own cloud bill directly, and a named forward-deployed engineer on every account to tune pipelines. CEO and co-founder Gabriel-James Safar argues the incumbent model of sending telemetry to a vendor cloud stopped working once volumes grew and AI systems began writing and deploying code. Founded in Paris in 2024, Tsuga raised roughly $10M in seed funding in December 2025 and a $35M Series A on 23 June 2026 led by Singular, with General Catalyst, DST Global Partners, QuantumLight, Picus and Databricks Ventures participating. Named customers include Le Monde, Camunda, Buk, Black Forest Labs, Aikido, Believe, Clara and Hippocratic AI.
The platform-engineering or SRE lead at a company whose Datadog or Dynatrace bill is growing faster than its infrastructure, and who is already sampling or dropping telemetry to control it
Full-fidelity telemetry with unlimited retention at a flat per-GB rate, with the storage and compute cost landing transparently on your own cloud bill instead of a vendor markup
At a Glance
- Category
- Infrastructure & Cloud
- Pricing
- Usage-based, Flat rate, Contact for pricing
- Target Market
- Platform Engineering Leads, Site Reliability Engineers, CTOs, VPs of Infrastructure, Enterprise Developers
- Deployment
- Self-hosted, Hybrid, Multi-cloud
- Founded
- 2024
- Headquarters
- Paris, France
Key Features
- ✓Bring-your-own-cloud deployment
Deploys into your own cloud account via infrastructure-as-code, so telemetry never crosses into a vendor-operated environment
- ✓Customer-managed encryption keys
Data is encrypted with your own KMS keys in your own object storage, which is what makes the sovereignty claim enforceable rather than contractual
- ✓AI agent observability
Agent traces, prompt and token visibility, confidence metrics and agent call graphs sit alongside conventional application and infrastructure telemetry
- ✓Observability pipelines
Field-level transformation and routing let teams reshape, enrich and redirect telemetry before it is stored, cutting volume at the source
- ✓Flat per-gigabyte pricing with unlimited retention
One rate per GB ingested with no per-field or retention surcharge, which removes the incentive to sample data away
- ✓Sensitive data scanning
Detects secrets and regulated data inside telemetry streams before it is persisted, reducing the compliance blast radius of verbose logging
- ✓Named forward-deployed engineer
Every account gets a dedicated field engineer to tune pipelines, aligning the vendor with reducing your volume rather than growing it
Capabilities
Use Cases
- •Cutting an out-of-control observability bill
Replace a per-host, per-field SaaS contract with flat per-GB ingest so retention can be raised rather than cut to hit budget
- •Data-sovereign monitoring in the EU
Keep logs, metrics and traces inside an EU or sovereign cloud account under customer-held keys to satisfy residency requirements
- •Debugging production LLM agents
Trace an agent's call graph, prompts and token spend alongside the infrastructure telemetry from the same request to find where it failed
- •Eliminating telemetry sampling
Store full-fidelity data with unlimited retention so incident investigations are not blocked by the one trace that was sampled away
- •Redacting sensitive data before storage
Scan and transform telemetry in the pipeline so secrets and regulated fields never reach long-term storage in the first place
Ideal For
Best For
- ✓Platform teams whose observability spend has outgrown its value and who are sampling or dropping logs purely to control cost
- ✓Regulated or EU-based organisations that need telemetry to stay inside their own cloud perimeter under customer-managed encryption keys
- ✓Teams instrumenting LLM agents that need prompt, token and agent call-graph visibility alongside conventional application and infrastructure traces
- ✓Companies running on sovereign or non-US clouds where mainstream observability SaaS has no acceptable data-residency story
- ✓Engineering orgs that want unlimited retention and unlimited custom fields without per-field or per-host pricing penalties
Not Ideal For
- ✗Small teams without cloud-operations maturity — bring-your-own-cloud means you own the AWS account, the IAM, the KMS keys and the underlying storage and compute bill, which is real operational work a SaaS vendor would otherwise absorb
- ✗Buyers who need published, self-serve pricing to budget: Tsuga has no pricing page, and independent coverage puts average contract values in six figures with a forward-deployed-engineer model that implies large deals
- ✗Organisations wanting a mature marketplace of third-party integrations and community-contributed dashboards — Tsuga was founded in 2024 and its ecosystem is nowhere near Datadog's or Grafana's
- ✗Teams needing a proven multi-cloud deployment today: the documented under-two-hours infrastructure-as-code path is the AWS one, with Azure, GCP and sovereign clouds stated but less evidenced
Deployment
Market Analysis
Pros
- ✓The bring-your-own-cloud architecture solves data residency and cost markup in one move, rather than trading one against the other
- ✓Flat per-GB pricing with unlimited retention and fields directly attacks the pricing mechanics that make incumbent observability bills unpredictable
- ✓AI agent traces, prompts, tokens and call graphs are native to the platform rather than a separate LLM-observability product bolted alongside APM
- ✓Credible early traction for a 2024 company: millions in contracted ARR within six months of leaving stealth, with named customers including Le Monde, Camunda and Black Forest Labs
- ✓Investor set includes Databricks Ventures and General Catalyst, which is a meaningful signal in the data-infrastructure category specifically
Cons
- ✗Bring-your-own-cloud shifts real operational burden onto the customer — you own the account, the keys, the storage lifecycle and the underlying cloud bill
- ✗No published pricing and a forward-deployed-engineer model with reported six-figure average contract values means this is not accessible to smaller teams
- ✗Founded in 2024, so the integration ecosystem, community dashboards and third-party tooling are far behind Datadog and Grafana
- ✗Multi-cloud support is stated but only the AWS deployment path is documented in detail, so Azure, GCP and sovereign-cloud buyers should verify before committing
- ✗No G2, Capterra, TrustRadius or Hacker News presence yet, so there is no independent practitioner feedback to validate the cost and reliability claims
Pricing
Trial
$0
- ✓No credit card required
- ✓Full platform access during evaluation
Bring-your-own-cloud
Contact for pricing
- ✓Flat rate per GB ingested
- ✓Unlimited retention
- ✓Unlimited fields
- ✓No markup on storage or compute
- ✓Named forward-deployed engineer
Tsuga publishes no rate card — the model is a flat charge per gigabyte ingested with unlimited retention and fields, on top of which you pay your own cloud provider directly for the storage and compute, since the platform runs in your account. Independent coverage reports six-figure average contract values, so this is an enterprise purchase; a no-credit-card trial is offered but the actual per-GB number requires a sales conversation.
Security & Compliance
Sources
This page was written from 5 sources, 4 on domains other than tsuga.com.
- 1.tsuga.com — tsuga.comvendor
- 2.pathfounders.com — tsuga raises 35m series a as ai strains legacy
- 3.thesaasnews.com — tsuga raises 35 million series a
- 4.siliconrepublic.com — paris based ai company tsuga raises 35m series a funding gro
- 5.eu-startups.com — french observability startup tsuga lands e30 million to expa
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
Coralogix
AI-native observability that queries logs, metrics and traces in place — index-free, in your own S3 bucket
OpenObserve
Open-source, Rust-based observability on object storage — logs, metrics, traces and LLM telemetry in one binary
SkyPilot
Run and scale AI workloads across Kubernetes, Slurm and 25+ clouds from one interface
MyDecisive
Open-source, OpenTelemetry-native telemetry control plane that acts on incidents instead of just charting them