Firebolt
firebolt.ioFirebolt is an analytical database, in the data warehouse category, that runs SQL queries over HTTP, an official MCP server and a CLI.
Can an AI agent use Firebolt? No. Our agent couldn't get into Firebolt on its own.
From finding it to leaving
What an agent managed without a person, at each step.
- On its own
Find
Docs it can act on: llms.txt, an OpenAPI and an MCP server. 544 doc pages read.
- Stopped
Get in
Stopped hereCan't sign up. The sign-up call was refused with a message that names no reason.
- Not reached
Use
Tasks need an account.
- Not reached
Pay
Buying needs an account.
- Not reached
Leave
Nothing to cancel or delete.
Getting in
None of the 2 ways in got our agent in on its own; one got partway.
Where it stopped
Email and password
The sign-up call was refused with a message that names no reason.
- Both kinds of agentOn a server, or in its own browserNoSigning up needs a person: the form refused our agent's browser.
- Signed agentKnown to CloudflarePartlyThe sign-up page opens. The check on the form still applies.
Every way in
| Way in | What happened |
|---|---|
| Runs locally, no account | Partway |
| Email and password | Refused |
What an agent can do with Firebolt
No task was tried: our agent couldn't get into Firebolt.
Core tasks Firebolt offers an agent
- offered, not tried
- Run a database queryAPIMCPCLI
- Create a databaseAPICLI
- Create a tableAPICLI
- List databases and read their schemaAPIMCPCLI
- Add or change a table fieldAPICLI
- Sync external data into a tableAPICLI
- Run a vector searchAPIMCPCLI
- Run a full-text searchAPICLI
- Query a database over a REST APIAPI
- Build the documented vector index on a default tableThe engine says this is not yet supported; the docs do not say so.
- Pay for the hosted serviceBilling runs only through AWS Marketplace after the trial.
13 more tasks it offers
- Create a table viewAPICLI
- Export data to a warehouseAPIMCPCLI
- Stream database changesAPI
- Resize database computeAPIMCPCLI
- Delete a databaseAPICLI
- Delete a recordAPICLI
- Create a database userAPICLI
- Set row-level security policiesAPI
- Configure SAML SSOWeb app
- Restrict access by IP addressAPIMCPCLI
- Read the audit logWeb app
- Create a sub-accountAPIMCPCLI
- Enable two-factor authenticationWeb app
Surface audit: 64 documented operations, 44 run, 37 passed, 6 not as documented, 18 blocked (sandbox or plan limits), 3 left out to stay safe.
Self-hosted edition
L4Runs on its ownFirebolt also ships an edition you run yourself. Our agent installed it on its own server in about 6 minutes and drove it the same way, doing 100% of the core work.
How an agent connects
MCP, API, OpenAPI, CLI, SDK and llms.txt found.
MCP server
8 testedOfficial server. Local binary; 4 tools in Core mode.
9 tools across its docs
API
17 testedSQL over HTTP; managed REST via service-account token.
22 operations across its docs
OpenAPI
Yes3.0.0, one operation (POST /query)
CLI
18 testedfirebolt CLI 5.0.0 (client, server, profile, session)
19 commands across its docs
SDK
Listed, not testedPython, Node.js, Go, JDBC, .NET, Rust, SQLAlchemy.
7 methods across its docs
llms.txt
YesAgent payments
None foundThe checkout wasn't reached: Managed service (compute 1x S), $0.92 a call.
Docs
No gaps in what we readCovers getting started, auth, error codes and code examples.
544 of 693 pages read
Tips for agent builders
- No account needed for the open-source edition: fetch packdb-linux-amd64.tar.gz from github.com/firebolt-db/firebolt-core releases and run 'firebolt server start --http-port 3473'; in a container or sandbox with a low memlock limit set execution.io_uring: off in <data-dir>/config.yaml.
- Send one SQL statement per HTTP request to http://<host>:3473/?output_format=JSON_Compact&database=<db>; a multi-statement body is refused with a clear message.
- Run the official MCP server with --core-url http://localhost:3473 --transport streamable-http; call firebolt_docs_overview first for the docs_proof, and pass any value as 'account' to firebolt_query on Core.
- On Core 5.0.0-pre, rank rows by vector_cosine_distance when the vector index does not build.
- Check the database name yourself before writing: confirm it in information_schema.catalogs.
How Firebolt compares
In Data warehouses
2 of the 7 products we drove in Data warehouses are higher; 4 others share L1.
| Product | Level | Find to leave |
|---|---|---|
| L4 | Find: On its own, Get in: On its own, Use: On its own, Pay: Card checkout, Leave: On its own. | |
| L4 | Find: On its own, Get in: On its own, Use: On its own, Pay: Not reached, Leave: On its own. | |
| L1 | Find: On its own, Get in: Stopped, Use: Not reached, Pay: Not reached, Leave: Not reached. | |
| L1 | Find: On its own, Get in: Stopped, Use: Not reached, Pay: Not reached, Leave: Not reached. | |
| L1 | Find: On its own, Get in: Stopped, Use: Not reached, Pay: Not reached, Leave: Not reached. | |
| L1 | Find: On its own, Get in: Stopped, Use: Not reached, Pay: Not reached, Leave: Not reached. | |
| L1 | Find: On its own, Get in: Stopped, Use: Not reached, Pay: Not reached, Leave: Not reached. |
Sorted by level, then work done, then time to the first call. How we rank
For the team behind Firebolt
Agents get stuck at Get in. Prove Firebolt is yours, then see where each issue is, read your safety findings and ask for a free drive after a fix. How to fix each one comes with the Audit.
7 things to fix for agents
Agent blockers: 1 foundAgent surfaces: 6 issues foundSafety review: 2 issues found
- FindClear
- Get inBlocks agents
- UseNot reached
- PayNot reached
- LeaveNot reached
On our standard
24 criteria: against other data warehouses products where we measured 5 or more, otherwise against every product we measured.
- Typed, tight input schemasAhead of most 50 of 65 others (every product) lower, 15 the same.
- Agents pick the right toolAhead of most 46 of 65 others (every product) lower, 19 the same.
- Docs an agent can navigateAhead of most 4 of 6 others (data warehouses products) lower, 2 the same.
- A tool list that fitsAhead of most 33 of 65 others (every product) lower, 32 the same.
- Spend an agent can see and capLevel with most 3 of 5 others (data warehouses products) the same, 2 higher.
- Structured resultsBehind most 22 of 36 others (every product) higher, 14 the same.
- Truthful annotationsBehind most 48 of 65 others (every product) higher, 17 the same.
- Versions an agent can pinThe lowest result All 4 others (data warehouses products) higher.
All 24 criteria
- The best way in is built for agentsIn the middle 3 of 6 others (data warehouses products) lower, 3 the same.
- A request id on every answerAhead of most 3 of 5 others (data warehouses products) lower, 1 the same, 1 higher.
- Prices an agent can readLevel with most 4 of 6 others (data warehouses products) the same, 2 lower.
- Sign-in for agents acting for a personLevel with most 4 of 6 others (data warehouses products) the same, 2 lower.
- Fast to the first callLevel with most 3 of 4 others (data warehouses products) the same, 1 lower.
- Descriptions that say when to use the toolIn the middle 20 of 65 others (every product) lower, 30 the same, 15 higher.
- Tool names that sort themselvesLevel with most 61 of 65 others (every product) the same, 4 lower.
- Every way in works for an agentIn the middle 2 of 6 others (data warehouses products) lower, 2 the same, 2 higher.
- Real status codesThe same as all 5 others (data warehouses products).
- Errors a program can readIn the middle 1 of 4 others (data warehouses products) lower, 2 the same, 1 higher.
- Errors say what to do nextIn the middle 1 of 4 others (data warehouses products) lower, 2 the same, 1 higher.
- Export before leavingLevel with most 4 of 6 others (data warehouses products) the same, 1 lower, 1 higher.
- WebMCP tools on the web appThe same as all 5 others (data warehouses products).
- Pages fit one fetchLevel with most 5 of 6 others (data warehouses products) the same, 1 higher.
- A path written for agentsIn the middle 1 of 6 others (data warehouses products) lower, 2 the same, 3 higher.
- Results sized for a context windowBehind most 16 of 31 others (every product) higher, 11 the same, 4 lower.
Where Firebolt fits
- Category
- Wall
- Bot checkskept our agent out
- Google reCAPTCHA bot checkthe check, and every product we met it at
- Let signed AI agents past your bot checkfix guide
- Rivals
Found an error? Tell us on this page or read how corrections work.
Questions people ask about Firebolt
- Can an AI agent sign up for Firebolt on its own?
- No. Sign-up: blocked. A bot check from Google reCAPTCHA stopped our agent there. The form at go.firebolt.io/signup (name, company e-mail, password. Ensure your details are correct and try again.' for the agent's inbox and again for the agent's inbox. No visible challenge and nothing else asked; no sign-up API documented. Asked for a role, a goal and a current platform; our agent answered.
- Does Firebolt have an MCP server, and can an agent connect to it?
- Yes. Firebolt has an official MCP server. The server runs on the agent's own machine and lists 4 tools. A new agent can't get in on its own: in the drive of 28 Oct 2026, our agent stopped at sign up (it was blocked). Through the MCP server, 0 of our agent's 6 attempts worked.
Cite this page
The page as a reference, in plain text, APA and BibTeX.
Cite this
AgentDrives, “Firebolt: AI agent test drive”, 28 October 2026, https://agentdrives.com/tools/firebolt. CC BY 4.0.
AgentDrives. (2026, October 28). Firebolt: AI agent test drive. https://agentdrives.com/tools/firebolt
@misc{agentdrives_firebolt_2026,
author = {{AgentDrives}},
title = {Firebolt: AI agent test drive},
year = {2026},
month = oct,
url = {https://agentdrives.com/tools/firebolt},
note = {Agent test drive, CC BY 4.0},
}Something wrong on this page?
Tell us what's wrong. We check it against the drive and correct the page, with the date.