Hosted one-shot code execution: send a snippet plus optional files, get stdout, stderr, exit...
Scored across 2 tools
The two tools target distinct input types: execute-code runs Python/JS/TS code, while execute-command runs shell commands. However, overlap exists because shell commands can invoke language interpreters, creating potential for misselection.
Both tool names follow the same verb_noun pattern: execute-code and execute-command. The convention is consistent and predictable.
With only 2 tools, the set feels thin for a sandbox execution service. While each tool serves a distinct core function, the count is borderline minimal rather than well-scoped.
The tools cover the primary operations of running code and shell commands in a one-shot sandbox. Minor gaps exist, such as explicit file upload or environment configuration, but agents can work around these using shell commands.
Runs Python, JavaScript or TypeScript code in an isolated one-shot sandbox and returns stdout, stderr, exit code and artifacts (Daytona CodeRunRequest/ExecuteResponse compatible) Price: $0.002 a call.
| Name | Required | Description | Default |
|---|---|---|---|
| env | No | Environment variables; keys must match [A-Za-z_][A-Za-z0-9_]* (Daytona's rule) | |
| argv | No | Command-line arguments for the program (sys.argv[1:] in Python, process.argv.slice(2) in Node) | |
| code | Yes | The code to run (up to 1 MB of UTF-8) | |
| files | No | Text files written into the work directory next to the program (up to 50, 1 MB each, 4 MB total); the entry point main.py/main.js/main.ts is reserved | |
| outputs | No | Relative paths to read back after the run (up to 20); up to 1 MB each comes back base64 in outputFiles, larger in tooLargeOutputs, absent in missingOutputs | |
| timeout | No | Run timeout in seconds, 1-120 (default 60); a timed-out run answers exitCode 124 | |
| language | Yes | "python", "javascript" or "typescript" |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description carries the full burden. It usefully discloses isolation ('isolated one-shot sandbox'), the return payload, and cost ($0.002 a call), which is genuinely helpful context. However, it omits key behavior for a code-execution tool: sandbox limitations (network egress, filesystem persistence), auth/permission needs, and rate limits.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Effectively one front-loaded sentence that names the capability, the return values, and the price without filler. The '(Daytona CodeRunRequest/ExecuteResponse compatible)' fragment is jargon that adds little for an agent, and the title largely repeats the description.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a 7-parameter execution tool with no output schema, the description correctly enumerates the return values so return-format documentation is not needed. It is still incomplete on operational constraints (network, persistence, permissions) and gives no routing signal against execute-command.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, so the schema already documents all 7 parameters in detail (env key rules, argv semantics, file limits, output readback behavior, timeout range). The description adds no parameter-level meaning beyond what the schema provides, making the baseline 3 correct.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
States a specific verb and resource ('Runs Python, JavaScript or TypeScript code'), names the languages, and says what comes back (stdout, stderr, exit code, artifacts). It does not distinguish itself from the sibling execute-command, so the agent cannot tell the two apart from the description alone.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
There is no explicit when-to-use guidance and no comparison with the sibling execute-command. The phrase 'one-shot sandbox' hints at ephemerality, but the description never states when this tool is preferable to execute-command.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Runs a shell command in an isolated one-shot sandbox and returns stdout, stderr and the exit code (Daytona ExecuteRequest/ExecuteResponse compatible) Price: $0.002 a call.
| Name | Required | Description | Default |
|---|---|---|---|
| cwd | No | Working directory, relative to the run's work directory | |
| env | No | Environment variables; keys must match [A-Za-z_][A-Za-z0-9_]* (Daytona's rule) | |
| files | No | Text files written into the work directory before the run (up to 50, 1 MB each, 4 MB total) | |
| command | Yes | The shell command to run (up to 64 KB) | |
| outputs | No | Relative paths to read back after the run (up to 20); up to 1 MB each comes back base64 in outputFiles, larger in tooLargeOutputs, absent in missingOutputs | |
| timeout | No | Run timeout in seconds, 1-120 (default 60); a timed-out run answers exitCode 124 |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description carries the full burden, and it does disclose three useful traits: the sandbox is isolated, it is one-shot (implying no persistence between calls), and each call costs $0.002. It does not disclose network access, filesystem persistence guarantees, or any auth/permission requirements, so meaningful gaps remain.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
A single front-loaded sentence covering action, scope, return values and price with no padding. The 'Daytona ExecuteRequest/ExecuteResponse compatible' clause is the only slightly jargon-heavy element, but it is compact and does earn its place for compatibility expectations.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a 6-parameter tool with nested objects, no annotations and no output schema, the description adequately names the return values (stdout, stderr, exit code) and the one-shot isolation model, while the schema fully covers parameter details. It leaves the outputFiles/tooLargeOutputs/missingOutputs structure and any size limits to the schema, which is acceptable given the coverage there.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, so the schema already documents cwd, env, files, outputs and timeout in detail; baseline 3 applies. The reference to 'Daytona ExecuteRequest/ExecuteResponse compatible' points the agent at external docs but adds no semantics beyond what the schema already states.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
States a specific verb and resource ('Runs a shell command') plus the scope ('isolated one-shot sandbox') and the return payload (stdout, stderr, exit code). It is distinguishable from execute-code by the 'shell command' framing, but it never names or contrasts the sibling, so it stops short of a 5.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
There is no when-to-use guidance at all: no statement of when a shell command is preferable to execute-code, no prerequisites, no note on whether the sandbox is stateful across calls. The only steer is the implicit 'shell' vs 'code' wording and the per-call price, which is cost context rather than usage guidance.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
execute-codeexecute-commandExecute code in 8 languages (Python, JS, TS, Go, Java, C++, C, Bash) in gVisor sandboxes.
Run Python code in a secure sandbox without local setup. Declare inline dependencies and execute s…
Hosted Python sandbox for package runtime verification: pip install and usable-import evidence.
Build Apps and run code in 30 languages — sandboxed, with persistent sessions for agent loops.
Add one secure layer between your agents and this server.
